KR20150050486A - Image processing apparatus, image processing method and program thereof - Google Patents

Abstract

The present invention provides an image processing device and the like that can generate a composite image corresponding to a desired focusing condition. In a smartphone (1), an edge detecting means (107) detects an edge as a feature from a plurality of input images taken with different focusing distances, and detects the intensity of the edge as a feature value. A depth estimating unit (111) then estimates the depth of a target pixel, which is information representing which of the input images is in focus at, with respect to the target pixel, by using the edge intensity detected by the edge detecting means (107). A depth map generating unit (113) then generates a depth map based on the estimation results by the depth estimating unit (111).

Description

TECHNICAL FIELD The present invention relates to an image processing apparatus, an image processing method,

The present invention relates to an image processing apparatus and the like that perform image processing.

When an object of interest is picked up by an image pickup device such as a camera, scene information including the background surrounding the object is read, and an image related to the scene information is output. At this time, depending on how the focus on the subject and the background is adjusted, the impression of the output image changes.

Here, when it is desired to output an image of a target object emphasized, the depth of field is set to be shallow for focusing on a subject of a near-distant view, while focusing on a background of a distant view There is a way. In order to reduce the depth of field, it is necessary to perform close-up imaging, use a telephoto lens, or open the diaphragm. It is difficult for the general person to perform these measures or to provide hardware capable of performing these measures. Further, in a mobile phone having a camera function such as a smart phone, it is mainstream to output a pan-focus image which is very limited in hardware, deep in depth of field, Thus, it is difficult to acquire an image with a shallow depth of field on the hardware of these widely used devices.

Here, as shown in Patent Document 1 or the like, when an image pickup operation is performed in order to make an image output emphasizing the object of interest even with a camera in which it is difficult to make the depth of field shallow, a focused image in which a focused object image is focused, (So-called soft focus image) around the outline of the subject is obtained by combining the focused image and the non-focused image automatically and continuously capturing the unfocused non-focused image automatically Technology has been proposed.

Japanese Laid-Open Patent Publication No. 2003-8966

However, when a plurality of subjects are included in the photographed image obtained by imaging, the other subject is focused (focused) on one of the two subjects but acquires a composite image of non-focused (visible) There is a case where the user wants to acquire a composite image of non-focusing although one subject is focused on the opposite case, that is, the other subject. In such conventional image processing apparatuses and the like, there is a problem that it is difficult to obtain a synthesized image corresponding to the desired focusing state.

An object of the present invention is to provide an image processing apparatus and the like capable of generating a composite image corresponding to a desired focusing state.

In order to solve the above problems, the present invention takes the following measures. In the following description of the embodiments for carrying out the invention described below and the reference numerals used in the drawings, reference should be made to the accompanying drawings, but the constituent elements of the present invention are not limited to those given the corresponding numerals.

According to a first aspect of the present invention,

(Feature amount, feature point) from a plurality of input images (a plurality of captured images picked up by the image pickup section 50) (step A11 in Figs. 4 and 15, step J7 in Fig. 19) (The processing section 10 in Fig. 1: an edge detection section 107, for example)

(Step A17 in Fig. 15, step F17 in Fig. 15, step J27 in Fig. 19) based on the detection result of the feature detecting means, the depth map generating means (113)).

Further, as another invention,

(Feature amount, feature point) from a plurality of input images (a plurality of captured images picked up by the image pickup section 50) (step A11 in Figs. 4 and 15, step J7 in Fig. 19) ,

(Step A17 in Fig. 4, step F17 in Fig. 15, step J27 in Fig. 19) to generate a depth map based on the detection result of the feature.

Further, as another invention,

The computer (smart phone 1, image processing device, other information processing device in Fig. 1)

(A feature amount, a feature point) from a plurality of input images (a plurality of captured images picked up by the image pickup section 50) (step A11 in Figs. 4 and 15, step J7 in Fig. )and,

(A step A17 in Fig. 4, a step F17 in Fig. 15, and a step J27 in Fig. 19) for generating a depth map based on the detection result of the feature detecting step (81).

According to the first invention or the like, a predetermined feature is detected from a plurality of input images, and a depth map is generated based on the detected feature. Based on the features detected from the plurality of input images, a depth map can be generated.

The depth map is information for specifying which input image is focused (i.e., the focused input image) with respect to the pixels of each coordinate, and the coordinates of the focused pixel for each input image (I.e., a region in which information is recorded). With the depth map generated in this manner, a composite image corresponding to a desired focusing state can be generated. As the " synthesized image corresponding to the desired focusing state ", for example, a predetermined subject is focused, but the other subject is a synthesized image that is unfocused or a composite image Quot;).

It is also possible to generate another synthesized image by modifying the subject to be focused or changing the degree of non-focusing after the synthesized image corresponding to the desired focusing state is once created by the depth map.

In addition, it is possible to generate a composite image corresponding to a desired focusing state by storing the input image used when generating the depth map and the depth map. Therefore, it is not necessary to generate a plurality of composite images at the time of image capture, The amount of data can be reduced, and the burden of managing image files and the like can be reduced.

Particularly, the present invention can be suitably applied to an image processing apparatus that generates a pan-focus image in which the depth of field is deep and is focused by focusing from the root to the far-end, such as a mobile phone having a camera function such as a smart phone.

Further, as a second invention,

In the image processing apparatus of the first invention,

The feature detecting means has feature amount detecting means (the processing portion 10 of Fig. 1) for detecting a predetermined feature amount (edge strength, DoG (Difference of Gaussian) value, background degree, etc.) from the plurality of input images ,

A feature amount detecting unit that detects a feature amount of each of the plurality of input images based on the feature amount detected by the feature amount detecting unit; (The depth estimating unit 111 shown in Fig. 1) for estimating the depth (step A15 in Figs. 4 and 15)

The depth map generating means may constitute an image processing apparatus that generates the depth map based on the estimation result of the depth estimation means (step A17 in Fig. 4, step F17 in Fig. 15).

According to the second aspect of the present invention, the feature amount detected from the plurality of input images is used to estimate a depth indicating which of the plurality of input images is focused on the target pixel. The depth map can be generated based on the estimation result as long as it is possible to estimate the depth indicating which of the plurality of input images is focused on the target pixel. The target pixel may be all the pixels in the image, or may be a pixel included in a part of the image.

In this case, as another invention,

The feature amount detection means has edge detection means (the edge detection portion 107 in Fig. 1) for detecting an edge as a feature from the plurality of input images,

The depth estimating means may use the intensity of the edge detected by the edge detecting means as the feature amount to estimate the depth with respect to the target pixel.

According to this other invention, an edge is detected as a feature from a plurality of input images, and the depth of the target pixel is estimated using the intensity of the detected edge as a feature amount. According to the discovery of the inventor of the present invention, the intensity of the edge detected from the input image has a correlation with the degree of focusing at the focal length. Thus, by using the intensity of the edge as the characteristic amount, it is possible to appropriately estimate the input image focused on the target pixel.

Further, as a third invention,

An image processing apparatus according to a second aspect of the present invention,

Wherein the depth estimation means calculates a depth parameter value for each of the plurality of input images detected by the feature amount detection means by using a feature amount of each of the plurality of input images, (Cost value) of the target pixel and a second input image that is focused on the vicinity of the target pixel and the first input image that is focused on the target pixel are different from each other (A score calculation process for maximizing or minimizing the value of the energy function: steps C1 to C7 in Fig. 7) using a second type parameter value (penalty value) The image processing apparatus may be configured to determine an input image that has been focused on.

According to the third aspect of the present invention, it is possible to appropriately determine an input image focused on a target pixel by performing predetermined appropriate arithmetic processing using the first kind parameter value and the second kind parameter value.

In this case, as another invention,

In the above-mentioned appropriate calculation processing, a predetermined score value is calculated by subtracting (subtracting) the second kind parameter value from the first kind parameter value with respect to the target pixel,

The depth estimation unit may configure the image processing apparatus to determine, for the target pixel, that the input image satisfying the predetermined high value condition is the focused image (step C5 in Fig. 7) .

According to this other invention, in the arithmetic processing as appropriate, the score value is calculated by subtracting the second kind parameter value from the first kind parameter value with respect to the target pixel, and the score value is set to a predetermined high value The input image that satisfies the condition is determined as the focused input image. Thus, it is possible to suppress an input image having a large feature quantity as much as possible for a target pixel while preventing an input image from being changed as much as possible in a nearby pixel. By performing such a score calculation process, it is possible to correctly determine an input image focused on a target pixel.

Further, as another invention,

In the above-mentioned appropriate arithmetic processing, a predetermined score value is calculated for the target pixel by adding the second kind parameter value to the first kind parameter value,

The depth estimation unit may constitute an image processing apparatus for determining, for the target pixel, an input image in which the score value satisfies a predetermined low value condition as an focused image.

According to this other invention, in the arithmetic processing as appropriate, the score value is calculated by adding the second-type parameter value to the first-type parameter value for the target pixel, and the score value is set to a predetermined low value The input image that satisfies the condition is determined as the focused input image. This makes it possible to correctly determine an input image focused on a target pixel.

Further, as another invention,

The larger the difference between the depth in the case where the first input image is focused and the depth in the case where the second input image is focused in the appropriate arithmetic processing, (Penalty value table 89 in FIG. 3; P ₁ < P ₂ ), and the score value is calculated.

According to this other invention, as the difference between the depth in the case where the first input image is focused and the depth in the case where the second input image is focused is larger in the arithmetic processing as appropriate, To calculate a score value. Thereby, it is possible to calculate the score value so that an input image whose focus is as close as possible to the pixel nearest the target pixel is selected.

Further, as another invention,

The image processing apparatus may be configured to calculate the score value using the second type parameter values individually determined in association with the combination of the two input images among the plurality of input images good.

According to this other invention, the score value is calculated by using the second kind parameter value set corresponding to the combination of the two input images among the plurality of input images in an appropriate arithmetic processing. For example, when the focus of the first input image and the focus of the second input image are greatly different from each other, by setting a large value as the second type parameter, an input image in which an input image with a large difference in focus is focused It is possible to make the determination difficult.

Further, as another invention,

(The value of the energy function of Expression (2) and Expression (3)) with respect to the line constituting the input image with respect to the pixel on the line, Step C3),

The depth estimation unit may configure the image processing apparatus to determine an input image focused on the line constituting the input image based on the score value (step C5 in Fig. 7).

According to this other invention, in the arithmetic processing appropriately, the score value is calculated for the pixels on the line constituting the input image, and for the pixels on the line constituting the input image, the input image focused on the basis of the score value . Thus, the calculation for the depth estimation can be facilitated.

Further, as a fourth invention,

An image processing apparatus according to a third aspect of the present invention,

(10) of the object shaking area detecting means for detecting a shaking area of the subject from the plurality of input images,

Wherein at least one of the first type parameter value and the second type parameter value is used in an area other than the subject shake area for the subject shake area detected by the subject shake area detection device The image processing apparatus may be configured to perform calculation with a value different from the parameter value.

According to the fourth invention, a subject shake area is detected from a plurality of input images. Then, in the arithmetic processing appropriately, at least one of the first kind parameter value and the second kind parameter value is set to a value different from the parameter value used in the area other than the subject shake area, with respect to the subject shake area . Thereby, it is possible to appropriately determine the focused input image with respect to the pixels included in the subject shake area.

Further, as another invention,

(The contrast correcting unit 105 in Fig. 1) for performing contrast correction (step A9 in Fig. 4 and Fig. 15) on the plurality of input images,

The characteristic amount detecting means may constitute an image processing apparatus for detecting the characteristic amount from the plurality of input images corrected by the contrast correcting means (step A11 in FIG. 4 and FIG. 15).

According to this other invention, it is possible to detect the feature amount appropriately reflecting the feature of the input image by performing the contrast correction on the plurality of input images and detecting the feature amount from the plurality of contrast-corrected input images.

Further, as a fifth invention,

An image processing apparatus according to any one of the second to fourth aspects,

(Focus separation determination unit 109 in Fig. 1) that determines whether the focal points of the plurality of input images are appropriately separated (focus separation determination processing in Fig. 6)

The depth estimation unit may configure the image processing apparatus to estimate the depth with respect to the target pixel using the input image in which the determination result of the focus separation determination unit is affirmative (step B7 in Fig. 6) .

According to the fifth invention, it is determined whether or not the foci of a plurality of input images are appropriately separated, and the depth of the target pixel is estimated using the input image whose determination result is affirmative. If the focal point of the input image is not properly separated, there arises a problem that the target object and other background are not properly separated. Here, the depth of the input image is estimated with the focus distance appropriately changed, and the input image in which the focus of the input image is properly separated.

In this case, as another invention,

The focal separation determination means calculates correlation values (Step B1 in FIG. 6) for calculating correlation values between different images for an image corresponding to each of the plurality of input images whose feature amounts are detected by the feature amount detection means (The processing unit 10 in Fig. 1), and determines that the input image corresponding to the image in which the correlation value calculated by the correlation calculating unit satisfies the predetermined low threshold value condition is an input image in which the focus is properly separated (Step B5 in Fig. 6), an image processing apparatus may be configured.

According to this other invention, a correlation value between different images is calculated for an image corresponding to each of a plurality of input images whose feature amounts are detected by the feature amount detecting means, and the calculated correlation value satisfies a predetermined low threshold value condition Is determined as an input image in which the focus is appropriately separated. If the focal point of the input image is appropriately separated, the correlation value between the other images in which the feature amount is detected will be a somewhat smaller value. This makes it possible to reliably determine whether or not the focus of the input image is appropriately separated by using the correlation value between the other images as the index value.

Further, as another invention,

Wherein the focus separation determination means determines whether or not the number or the ratio of the features satisfies a predetermined high threshold value for the image corresponding to each of the plurality of input images for which the feature amount is detected by the feature amount detection means B3) The image processing apparatus may be configured to determine the input image as an input image in which the focus is properly separated (step B5 in Fig. 6).

According to this other invention, for an image corresponding to each of a plurality of input images whose feature quantities have been detected by the feature quantity detecting means, an input image satisfying a predetermined high threshold value in the number or ratio of the features, As the input image. If the focus of the input image is appropriately separated, the number and the ratio of the features will be somewhat large. Therefore, it is possible to reliably determine whether or not the focus of the input image is appropriately separated by using the number or the ratio of the features as the index value.

Further, as a sixth invention,

An image processing apparatus according to any one of the second to fifth aspects,

The characteristic amount detecting means detects the characteristic amount from the reduced input image (the input image reducing unit 101 in Fig. 1) obtained by reducing the plurality of input images or the plurality of input images (step A11 in Fig. 4) ,

The depth map generating means may construct an image processing apparatus that generates a reduced depth map based on the estimation result of the depth estimation means (step A17 in Fig. 4).

According to the sixth aspect of the present invention, a feature quantity is detected from a plurality of input images or a reduced input image obtained by reducing a plurality of input images. Then, based on the estimation result of the depth, a reduced depth map is generated. The reduced depth map can be generated by applying the depth map generation method of the present invention to a plurality of input images or a reduced input image obtained by reducing a plurality of input images. Since the reduced depth map has a smaller capacity than the depth map, the memory capacity is reduced when it is stored in the memory.

Further, as a seventh invention,

An image processing apparatus according to a sixth aspect of the present invention,

(Depth map restoring unit 115 in Fig. 1) for restoring the reduced depth map (step A19 in Fig. 4)

(Step A21 in Fig. 4) of generating a depth map in which the depth map restored by the restoration means is corrected using a predetermined correction image (input image) The image processing apparatus may further comprise:

According to the seventh invention, the reduced depth map is restored using a predetermined restoration method. The reconstruction of the depth map in this way causes a problem such that irregularities occur in a portion which becomes the boundary of the depth or the position of the edge in the boundary portion of the object deviates. Here, a depth map is created by modifying the restoration depth map, which is the restored depth map, by using a predetermined correction image. Thus, the above problem can be solved and an appropriate depth map can be obtained.

Further, as an eighth invention,

In the image processing apparatus of the first invention,

Wherein the feature detecting means detects the feature points from the input image by using the captured images (captured images captured in a parallax system) captured at a plurality of viewpoints (viewpoints) at which parallax occurs, as the plurality of input images (Step J7 in Fig. 19). With the feature point detecting means (the processing section 10 in Fig. 1)

(Camera matrix: rotation matrix, translational motion, translational motion vector) of each of the plurality of captured images based on the feature points detected by the feature point detection means J23) viewpoint positional relationship estimation means (the processing section 10 in Fig. 1)

(Step J25 in Fig. 19) for estimating a three-dimensional position of a pixel constituting the plurality of input images by performing a predetermined estimation calculation (bundle adjustment) ),

The depth map generating means may configure the image processing apparatus to generate the depth map based on the estimation result of the viewpoint positional relation estimation means and the estimation result of the pixel position estimation means (Step J27 in Fig. 19) .

According to the eighth aspect of the present invention, a relative positional relationship between a point of time estimated based on a feature point detected from a captured image captured at a plurality of time points where parallax occurs, and a plurality of input images estimated by performing a predetermined estimation calculation The depth map can be appropriately generated based on the three-dimensional position of the pixel.

Further, as a ninth invention,

An image processing apparatus according to an eighth aspect of the present invention,

The viewpoint position relation estimating means estimates,

A feature point tracking unit (the processing unit 10 in Fig. 1) that tracks feature points detected by the feature point detection unit (step J11 in Fig. 19)

(Step J13 in Fig. 19) for extracting a starting point input image (base frame) which is an input image which is a starting point (base point) in the plurality of input images based on the tracking result of the minutia point tracking means (10)), wherein, based on the feature points detected by the feature point detection means, the relative positional relationship between the start point of each of the plurality of shot images with respect to the base point input image extracted by the base point input image extraction means The image processing apparatus according to the present invention may be configured.

According to the ninth aspect of the present invention, on the basis of the feature points detected with respect to the base input image extracted from the plurality of input images on the basis of the tracking results obtained by tracking the feature points, the relative positional relationship between the viewpoints of the plurality of captured images can be accurately estimated .

Further, as a tenth invention,

An image processing apparatus according to the eighth or ninth aspect of the present invention,

The display means (display portion 30 in Fig. 1)

(Guide arrows and guide diagrams in Fig. 22) for guiding the user to the moving method (moving direction, moving amount) of the magnetic device for causing a parallax to the plurality of captured images to the display means (The display control unit 127 of Fig. 1) for controlling the display of the image data.

According to the tenth aspect of the present invention, by displaying the guide display for guiding the user to the moving method of the magnetic device for generating the parallax in the plurality of captured images, the user can see how the magnetic device It is very convenient because you can see if it should be moved.

Further, as an eleventh invention,

(Inertial sensor 75 (acceleration sensor, gyro sensor), processing unit 10, and inertial measurement unit (see FIG. 1)) for detecting a movement amount (amount, (IMU), a GPS unit, and an inertial navigation system (INS)

The display control means may constitute an image processing apparatus for changing the display of the guide display (the display position of the guide arrow or the display position of the guide graphic form in Fig. 22) based on the detection result of the movement amount detecting means .

According to the eleventh invention, by changing the display of the guide display based on the detected moving amount of the magnetic device, it is possible to notify the user of the appropriate moving direction and moving distance of the magnetic device.

Further, as a twelfth invention,

An image processing apparatus according to any one of the first to eleventh aspects,

(Fore-focus image generating unit 119 in Fig. 1) (step A23 in Fig. 4 and Fig. 19) for generating a full focus image using the depth map and the plurality of input images The image processing apparatus may be further provided.

According to the twelfth aspect, by using the depth map and the plurality of input images, it is possible to generate a foreground image which is an image in a focused state for each of the pixels constituting the input image.

Further, as another invention,

(Fore-focus image generation unit 119 of FIG. 1) for generating a foreground image using the depth map generated by the depth map generation unit and the plurality of input images (step F19 of FIG. 15) )and,

(Depth map reducing section 121 in Fig. 1) for reducing the depth map generated by the depth map generating means by using a predetermined reduction method (step F21 in Fig. 15)

(Storage unit (80), image data set (83) in FIG. 1) for storing the reduced depth map, which is the reduced depth map, in correspondence with at least one of the plurality of input images and the foreground image , An image processing apparatus may be configured.

According to this other invention, a foreground image is generated by using the depth map generated by the depth map generating means and a plurality of input images. Then, the depth map generated by the depth map generating means is reduced using a predetermined reduction method, and the reduced depth map, which is a reduced depth map, is stored in correspondence with at least one of a plurality of input images and a foreground image. By storing the reduced depth map, the memory capacity can be reduced compared with the case of storing the depth map itself. Further, by memorizing in correspondence with a plurality of input images or fore-focus images, it is possible to subsequently modify the fore-focus image using these image information, or use these image information in other apparatuses.

Further, as another invention,

(Depth map restoring unit 115 in Fig. 1) for restoring the reduced depth map stored in the storage unit using a predetermined restoration method (step A47 in Fig. 15)

(Step A49 in Fig. 15) for correcting the restored depth map, which is the restored depth map, by using a predetermined correction image (depth map correction unit 117 in Fig. 1) Maybe.

According to this other invention, the reduced depth map stored in the storage means is restored using a predetermined restoration method. Then, the restoration depth map, which is the restored depth map, is corrected using a predetermined correction image. The reconstructed depth map causes a problem such that irregularities occur in a portion that becomes the boundary of the depth or that the position of the edge in the boundary portion of the object deviates. Here, the restoration depth map, which is the restored depth map, is corrected using a predetermined correction image. Thus, the above problem can be solved and an appropriate depth map can be obtained.

In this case, as another invention,

Wherein the correction image is at least one of a front-focus image stored in association with the storage means or the plurality of input images,

The correction means corrects the pixel value of the correction target image of the correction target pixel and the correction image of the vicinity pixel of the correction target pixel with the pixel value of the depth map of the vicinity pixel of the correction target pixel, Weighted averaging process (weighted averaging process according to the equation (3)) is performed with a higher weighting value as the pixel value of the pixel to be corrected becomes closer to the target pixel and the closer the distance between the correction target pixel and the neighboring pixel becomes , And the pixel value of the restoration depth map of the correction target pixel is corrected.

According to this other invention, as the correction image, at least one of the input image of the fore-focus image stored in association with the storage means or the plurality of input images is used. The pixel value of the correction image of the correction target pixel and the pixel value of the correction image of the neighboring pixel are set to the pixel value of the depth map of the vicinity pixel of the correction target pixel, The pixel value of the restoration depth map of the correction subject pixel is corrected by performing a weighted averaging process with a higher weighting value as the distance between the correction subject pixel and the adjacent pixel becomes closer . By this method, a depth map in which the boundary between the foreground and the background is clear can be obtained.

Further, as a thirteenth invention,

An image processing apparatus according to any one of the first to twelfth aspects,

The display means (display portion 30 in Fig. 1)

An input unit (input unit 20 in Fig. 1) for accepting a user operation,

A composite image obtained by applying a pixel value of a pixel in an input image corresponding to a specific depth to an input image corresponding to a depth other than the specific depth on the basis of the depth map in response to a user operation on the input means Further comprises synthesized image generating means (emphasis image generating section 123 in Fig. 1)

The display means may constitute an image processing apparatus for displaying a composite image generated by the composite image generating means.

According to the thirteenth aspect of the present invention, the pixel value of the pixel in the input image corresponding to the specific depth is calculated at a depth other than the specific depth, based on the depth map, corresponding to the user operation input from the input means for accepting the user operation And generates a composite image synthesized with the corresponding input image. And displays the generated composite image. This makes it possible to generate a synthesized image focused on a portion designated by the user, for example, and to view it to the user.

Further, as a fourteenth invention,

An image processing apparatus according to a twelfth aspect of the present invention,

The display means (display portion 30 in Fig. 1)

An input unit (input unit 20 in Fig. 1) for accepting a user operation,

(10) of the shading processing unit (10) for shading the foreground image, and corresponding to a user operation on the input unit, the shading processing unit (An emphasized image generating unit 123 in Fig. 1) for generating a synthesized image in which a pixel value of a pixel in the input image is combined with an image resulting from the shading process,

The display means may constitute an image processing apparatus which displays the composite image generated by the composite image generating means.

According to the fourteenth invention, the pixel value of the pixel in the input image corresponding to the specific depth is shaded with respect to the foreground image based on the depth map, corresponding to the user operation input from the input means for accepting the user operation A synthesized image synthesized with the processed image is generated, and the synthesized image thus generated is displayed. This makes it possible to generate a synthesized image focused on a portion designated by the user, for example, and to view it to the user.

In this case, as another invention,

Wherein the composite image generation means determines whether or not an area on the image specified by the user operation is an area in which the shading process has been performed and, if the determination result is affirmative, the degree of shading of the area is set to the predetermined shading degree May be changed to another degree of shading, and the shading processing may be performed again on the composite image.

According to this other invention, it is possible to change the degree of shading of the image once shaded by the shading processing on the area on the image designated by the user operation, and to generate the re-shaded composite image. This makes it possible, for example, to change (shine or weaken) the degree of shading of the portion of the synthesized image by the user tapping the already shaded portion.

Further, as a fifteenth invention,

In the image processing apparatus of the fourteenth aspect,

(An elliptical shape, a circular shape, or a rectangular shape in Fig. 25) for specifying the focusing target area that is a focusing target area by the user in correspondence with the user operation (icon operation, tap operation) (The display control unit 127 of Fig. 1) for displaying on the display means,

The synthesized image generating means may construct an image processing apparatus that generates the synthesized image with the depth of the pixel included in the designation figure in the input image as the specific depth based on the depth map.

According to the fifteenth aspect of the present invention, in response to the user's operation on the input means, the display means displays a designation figure for allowing the user to designate the focusing target area, which is the area to be focused on, The area to be focused can be designated.

Further, as a sixteenth invention,

An image processing apparatus according to the fifteenth aspect of the present invention.

Wherein the input means has a touch panel integrally formed with the display means,

(1) a pinch operation (pinch-out operation / pinch-in operation) with respect to the designation-purpose design pattern, (2) a plurality of (Long press time in the case of a long press operation) in which the tap operation has been continued with respect to the designation figure, and (3) (To enlarge / reduce the designation figure), the image processing apparatus may be configured.

According to the sixteenth invention, the size of the designation figure can be changed by a simple and easy user operation.

In addition, as a seventeenth invention,

The image processing apparatus according to the thirteenth or fourteenth aspect,

The input means has a touch panel (the touch panel 21 of Fig. 1) integrally formed with the display means,

The composite image generating means may constitute an image processing apparatus for generating the composite image with the depth determined based on the operation on the touch panel at the specific depth.

According to the seventeenth invention, a composite image can be generated by an operation on a touch panel integrally formed with the display means.

Further, as the eighteenth invention,

An image processing apparatus according to a seventeenth aspect of the present invention,

Wherein the display means displays a confirmation screen (a screen in which an image based on an output signal from the imaging element is displayed in real time)

(Storage unit 80 in Fig. 1) for storing a focusing target position, which is a position on a confirmation screen to be focused, corresponding to a tap operation on the touch panel on the confirmation screen,

(Imaging section (50) in Fig. 1) for performing imaging in accordance with establishment of a predetermined imaging condition after the focusing target position is stored in the storage means,

The feature detecting means detects the feature by using a plurality of captured images captured by the image capturing means as the plurality of input images,

The composite image generating means may constitute an image processing apparatus that generates the composite image with the depth corresponding to the focusing target position stored in the storage means at the specific depth.

According to the eighteenth aspect of the present invention, the user can designate the focusing target position on the confirmation screen. Then, the imaging is performed after that, and a composite image in which the depth corresponding to the focusing target position specified on the confirmation screen is set to a specific depth can be generated.

In this case, as another invention,

(1) detecting a predetermined imaging operation on the input means, (2) detecting a continuous tap operation for a predetermined period of time or more, and 3) It is also possible to configure an image processing apparatus that captures an image by establishing one of the detection of a plurality of tap operations by establishing the image capturing condition.

According to this other invention, it is possible to perform imaging by a simple and easy user operation after the focusing target position is stored by the tapping operation.

Further, as a nineteenth invention,

An image processing apparatus according to a seventeenth aspect of the present invention,

The display means displays a predetermined display image (input image)

(The processing unit 10 in Fig. 1) for setting a focusing target area that is a focusing target area in correspondence with the pinch operation on the touch panel on the display screen on which the display image is displayed,

Shading processing means (processing portion (10) in Fig. 1) for performing shading processing with respect to an area other than a focusing target area set by the focusing target area setting means with a predetermined degree of shading,

Wherein the shading processing means changes the shading degree in accordance with a pinch operation (pinch-out operation / pinch-in operation) on the focusing target area set by the focusing object area setting means (the shading degree is increased / ) Shading degree changing means (the processing section 10 in Fig. 1).

According to the nineteenth aspect of the invention, the user can designate the focusing target area, which is the focusing target area, by tapping the touch panel on the display screen on which the display image is displayed. Then, a shading process is performed with respect to an area other than the focusing target area with a predetermined degree of shading, so that a combined image in which the focusing area is emphasized can be generated.

Further, as a twentieth invention,

An image processing apparatus according to a nineteenth aspect of the present invention,

Wherein the shading degree changing means gradually increases the degree of shading in correspondence with the pinch-out operation for the focusing target area and decreases the degree of shading stepwise in accordance with the pinching-in operation on the focusing target area Processing apparatus may be configured.

According to the twentieth aspect, by the simple and easy user operation, the degree of shading for performing the shading processing on the area other than the focusing area can be changed.

Further, as a twenty-first invention,

An image processing apparatus according to any one of the first to twelfth aspects,

Wherein the plurality of input images are a plurality of captured images captured at different focal lengths,

The display means (display portion 30 in Fig. 1)

An input unit (input unit 20 in Fig. 1) for accepting a user operation,

(Focal length slider S1 in Fig. 27) for specifying the focal length range for the user to specify a focal length range designation image at a predetermined position of the display means (the display control unit 127)

A synthesizing unit configured to focus a pixel of a depth corresponding to a focal length range specified by a user operation on the image for specifying the focal length range and to generate a composite image in which other pixels are non- (The enhanced image generation unit 123 in Fig. 1)

The display means may constitute an image processing apparatus for displaying a composite image generated by the composite image generating means.

According to the twenty-first aspect of the present invention, a user creates a composite image by focusing on a pixel of a depth corresponding to a desired focal length range by non-focusing on the image for specifying the focal length range The synthesized image can be browsed.

Further, as a twenty-second invention,

An image processing apparatus according to any one of the first to twelfth aspects,

Wherein the plurality of input images are a plurality of captured images captured at different focal lengths,

The display means (display portion 30 in Fig. 1)

An input unit (input unit 20 in Fig. 1) for accepting a user operation,

A plurality of reference images based on the plurality of input images (a plurality of input images, a feature detection image in which a feature is detected for each of a plurality of input images, a contrast-corrected image obtained by performing contrast correction for each of a plurality of input images, (The display control unit 127 in Fig. 1) for performing control to sequentially display on the display means an image obtained by multiplying an input image of the input image and a corresponding feature detection image, ,

Wherein the depth map is used to focus a pixel at a depth corresponding to a focal length of an input image corresponding to a reference image designated by the user among the sequentially displayed reference images, (An enhanced image generating unit 123 in FIG. 1) for generating a combined image,

The display means may constitute an image processing apparatus for displaying a composite image generated by the composite image generating means.

According to the twenty-second aspect of the present invention, a user browses a reference image sequentially displayed to designate a reference image in which a desired subject is strongly displayed, and, based on the depth map, , The synthesized image can be browsed by focusing the pixels corresponding to the focal length of the input image corresponding to the reference image designated by the user and by making the other pixels non-focusing.

Further, as a twenty-third invention,

An image processing apparatus according to a twenty-second aspect of the present invention,

(An image obtained by multiplying an input image and an edge image by pixel units) obtained by multiplying the input image and a feature detection image (edge image) whose characteristics are detected by the feature detection unit, for each of the plurality of input images, (The processing unit 10 in Fig. 1) that generates the image data,

And the display control means may constitute an image processing apparatus that performs control to sequentially display the red image generated by the red image generation means as the reference image on the display means.

According to the twenty-third aspect of the present invention, an enemy image obtained by multiplying the input image and the detected feature detection image by feature is generated for each of the plurality of input images, and sequentially displayed as a reference image, So that the user can easily designate it.

Further, as a twenty-fourth invention,

An image processing apparatus according to a seventeenth aspect of the present invention,

Wherein the synthesized image generating means dynamically changes the specific depth based on a drag operation on the touch panel to dynamically generate the synthesized image,

The display means may constitute an image processing apparatus for displaying a synthesized image dynamically generated by the synthesized image generating means in accordance with the drag operation.

According to the twenty-fourth aspect of the present invention, a composite image can be generated and displayed in real time based on a drag operation of the user on the touch panel.

Further, as a twenty-fifth invention,

An image processing apparatus according to any one of the second to seventh aspects,

(Input unit 20 in Fig. 1) for accepting user operations,

Wherein the depth estimation means has depth depth estimation means (the processing portion (10) in FIG. 1) for predicting the depth for the pixels included in the designated region specified by the user operation,

The depth map generating means may be configured to include the depth map regenerating means (the processing portion 10 of FIG. 1) for regenerating the depth map based on the result of re-estimation of the depth by the depth estimating means good.

According to the twenty-fifth aspect of the present invention, when a user finds a portion where an unnatural display is made in an image, the user designates the portion so that re-estimation of the depth is performed on the portion, Based on the result of the above, a depth map in which an unnatural part is appropriately corrected can be obtained.

Further, as a twenty-sixth invention,

An image processing apparatus according to a twenty-fifth aspect of the present invention,

Wherein the depth estimating means estimates at least one of the first type parameter value and the second type parameter value with respect to the pixels included in the designation region from a parameter value used in the original depth estimation by the depth estimation means It is also possible to construct an image processing apparatus that changes the depth to another value and appropriately performs the arithmetic processing to refine the depth.

According to the twenty-sixth aspect of the invention, at least one of the first-type parameter value and the second-kind parameter value is set to a value different from the parameter value used by the depth estimation means in the original depth estimation It is possible to appropriately perform the arithmetic processing and to re-estimate the depth appropriately.

Further, as another invention,

Further comprising pixel specifying means for specifying a predetermined pixel in the foreground image (for example, a pixel in an area tapped by the user on the touch panel or a pixel in an area in which the edge intensity is equal to or higher than a predetermined value or lower than a predetermined value)

Wherein the synthesized image generating means generates the synthesized image by reducing the degree of shading for the pixel specified by the pixel specifying means in the foreground image to be smaller than the predetermined degree of shading or setting the degree of shading for the pixel specified by the pixel specifying means The image processing apparatus may be configured to perform the shading processing with respect to the pixels other than the pixels by making the shading degree larger than the predetermined shading degree.

When a pixel having a depth to be emphasized is to be emphasized on an image obtained by uniformly shading a foreground image, a phenomenon occurs in which an area that becomes an unnatural pixel value such as a contour near the boundary between focusing and non-focusing occurs. In order to avoid this phenomenon, the shading degree of the pixel specified by the pixel specifying means in the foreground image is made smaller than the predetermined shading degree, or the shading degree is set to a pixel other than the pixel specified by the pixel specifying means in the foreground image The degree of shading is made larger than the predetermined degree of shading. This makes it possible to avoid the above-described phenomenon and obtain an appropriate composite image.

Further, as another invention,

(For example, the depth designation operation window W1 in FIG. 17 or the focal length slider S1 in FIG. 27) for designating the specific depth to the user at a predetermined position of the display means And a display control means for controlling the display means,

The image processing apparatus may be configured to allow the user to specify the specific depth by using the input means in the depth designation operation window displayed on the display means.

According to this other invention, a control for displaying a depth designation operation window for designating a specific depth to a user is performed at a predetermined position of the display means. Then, the user designates the specific depth using the input means from the depth designation operation window displayed on the display means. Thereby, the user can designate an arbitrary specific depth from the operation window for specifying the depth. As a result, an emphasized image in which pixels of a specific depth are emphasized is generated and displayed on the display unit, so that the user can browse the combined image in which the pixels of the selected depth are highlighted.

Further, as another invention,

The input means has a touch panel (the touch panel 21 of Fig. 1) integrally formed with the display means,

The user may configure the image processing apparatus to designate the specific depth by a tap operation on the touch panel.

According to this other invention, the user can specify a specific depth by tapping the touch panel.

Further, as another invention,

The user operation is a tap operation,

The composite image generation means may select the specific depth based on the designation area by using the area on the image tapped by the tap operation as the designation area (step D1 in FIG. 8) D5), and an image processing apparatus may be configured.

According to this other invention, the user operation is a tap operation, and a region on an image tapped by the tap operation is designated as a designated region, and a specific depth is selected based on the designated region. This makes it possible to generate a composite image by selecting a specific depth for the tapped designated area in accordance with the user's tap operation, and to display the composite image on the display section.

Further, as another invention,

Further comprising a display control means for performing control to display either one of the focused image, the input image selected from the plurality of input images, the composite image, and the depth map as a display image on the display means,

And the display image may be referred to so that the user can designate the specific depth using the input means.

According to this other invention, the display control means controls the display means to display either one of the focused image, the input image selected from the plurality of input images, the composite image, and the depth map as the display image. Then, the user designates a specific depth by using the input means based on the display image that is displayed in a referential manner on the display means. Thereby, the user can browse the display image and designate the portion to be highlighted.

Further, as another invention,

(2) a depth map corresponding to a deepest or shallower pixel in the depth map among the pixels included in the designated area; (3) the depth corresponding to the pixel value closest to the average value of the pixel values of the pixels included in the designated area, (4) the pixel corresponding to the most corresponding depth on the depth map of the specified area The image processing apparatus may be configured to select one of the depths having a large area as the specific depth.

According to this other invention, (1) the depth corresponding to the pixel closest to the center position of the designated area, (2) the depth corresponding to the deepest or shallowest pixel in the depth map among the pixels included in the designated area, 3) the depth corresponding to the pixel value closest to the average value of the pixel values of the pixels included in the designated area, and (4) the depth of the designated area corresponding to the pixel having the largest depth corresponding to the depth map. A composite image can be generated by selecting one of them as a specific depth.

Further, as another invention,

Wherein the feature amount detection means has edge detection means (edge detection portion 107 in Fig. 1) for detecting an edge as a feature of the plurality of input images,

The input means has a touch panel (the touch panel 21 of Fig. 1) integrally formed with the display means,

The user operation is a drag operation,

Depth selecting means (the processing portion 10 in Fig. 1) for selecting one depth in accordance with the drag operation (Step E5 in Fig. 9)

(Step E6 in Fig. 9) of the direction of the drag operation by the user and the direction of the edge included in the designated area specified by the drag operation in the input image of the depth selected by the depth selection means (The processing section 10 in Fig. 1)

Wherein the depth estimating means estimates at least one of the first type parameter value and the second type parameter value with respect to the pixels included in the designation region from a parameter value used in the original depth estimation by the depth estimation means It is also possible to configure the image processing apparatus to perform the score value calculation processing by changing the value to another value and to re-estimate the depth.

According to this other invention, the feature amount detection means has edge detection means for detecting an edge as a feature of a plurality of input images. Then, one depth is selected in accordance with the dragging operation of the user to the touch panel, and the direction of the drag operation by the user and the direction of the edge included in the designated area specified by the drag operation of the selected depth image And calculates the degree of similarity. Then, the depth re-estimating means calculates at least one of the first-type parameter value and the second-type parameter value for the pixels included in the designated area designated by the drag operation as the parameter value used by the depth estimation means in the initial depth estimation The depth of the edge portion in the image can be appropriately re-estimated.

According to the present invention, there is provided an image processing apparatus and the like capable of generating a composite image corresponding to a desired focusing state.

1 is a block diagram showing an example of a functional configuration of a smartphone.
2 is a diagram showing an example of the data configuration of an image data set;
3 is a diagram showing an example of the table configuration of the penalty value table.
4 is a flowchart showing an example of a flow of image processing.
5 is a flow chart continued from FIG.
6 is a flowchart showing an example of the flow of the focus separation determination processing.
7 is a flowchart showing an example of the flow of the depth estimation processing.
8 is a flowchart showing an example of the flow of the enhanced image generation processing.
9 is a flowchart showing an example of the flow of the enhanced image modification processing.
10 (1) is an example of an input image, (2) is an example of an input image, (3) is an example of an edge image, and (4) is an example of an edge image.
11 (1) is an example of a depth map, and (2) is an example of a foreground image.
12 (1) is an example of a restoration depth map, and (2) is an example of a correction depth map.
13 (1) is an example of a shading image for a foreground image, (2) is an example of an enhanced image, (3) is an example of a shading image, and (4) is an example of an enhanced image.
14 (1) is an example of an enhanced image, and (2) is an example of a modified image for an enhanced image.
15 is a flowchart showing an example of the flow of the second image processing.
16 is a diagram showing an example of the table configuration of the penalty value table in the modified example.
17 is a diagram showing an example of an operation window for depth selection.
18 is a flowchart showing an example of the flow of the second enhanced image generation processing.
19 is a flowchart showing an example of the flow of the third image processing.
20 (1) is a diagram showing an example of a feature point detection result, (2) is an example of a feature point detection result, and (3) is an example of a feature point tracking result.
21 (1) is a diagram showing an example of the estimation result of the coordinates of the minutiae point at the frontal viewpoint, (2) is an example of the estimation result of the coordinates of minutiae points of the upper- Is an illustration showing an example of the estimation result of the coordinates of the minutiae point of the left eye point.
22 (1) is a diagram showing an example of a confirmation screen, (2) is an example of guide display, (3) is an example of guide display, (4) Fig.
23 is a flowchart showing an example of the flow of the fourth image processing.
24 is a flowchart showing an example of the flow of the third enhanced image generation processing.
FIG. 25 (1) is a diagram showing an example of a region designating figure, and (2) is an explanatory diagram of the size adjustment of a region designating figure.
Fig. 26 (1) is an explanatory diagram of the change of the degree of shading by the pinch-out operation, and (2) is an explanatory diagram of the change of the degree of shading by the pinch-in operation.
27 (1) is a view showing an example of a focal length slider, (2) is an example of a focal length slider, and (3) is an example of a focal length slider.
Fig. 28 is a diagram showing an example of the display of distinction of pixels. Fig.
29 is a diagram showing an example of the flow of the fourth enhanced image generation processing.
30 is a diagram showing an example of a recording medium.
31 is an explanatory diagram of the principle of generation of a depth map using a parallax system.
32 is a flowchart showing an example of a flow of a highlight clipping effect shading process.

Hereinafter, an example of a preferred embodiment to which the present invention is applied will be described with reference to the drawings. In this embodiment, an embodiment in which the present invention is applied to a smart phone, which is a type of mobile phone having a camera function, as an image processing apparatus will be described. However, it is needless to say that the embodiments to which the present invention is applicable are not limited to the embodiments described below.

Fig. 1 is a block diagram showing an example of the functional configuration of the smartphone 1 according to the present embodiment. The smartphone 1 includes a processing unit 10, an input unit 20, a display unit 30, a communication unit 40, an imaging unit 50, a cellular phone antenna 55, A wireless communication circuit unit 60, a clock unit 70, an inertial sensor 75, and a storage unit 80.

The processor 10 is a processor that controls each section of the smartphone 1 in accordance with various programs such as a system program stored in the storage unit 80. The processor 10 includes a CPU (Central Processing Unit) and a DSP Processor).

In the present embodiment, the processing unit 10 includes, as main functional units, an input image reducing unit 101, a aligning unit 103, a contrast correcting unit 105, an edge detecting unit 107, A depth map generating unit 113, a depth map restoring unit 115, a depth map correcting unit 117, a foreground image generating unit 119 A depth map reducing section 121, an emphasized image generating section 123, an emphasized image correcting section 125 and a display control section 127. The depth map reducing section 121, It goes without saying that these functional units are merely described as one embodiment, and that other functional units may be essential components.

The input unit 20 is an input device having a touch panel 21 and outputs a detection signal of a user operation to the touch panel 21 to the processing unit 10. [ The touch panel 21 functions as an input interface between the user and the smartphone 1. [

In the present embodiment, when the camera function of the smartphone 1 is selected through the input unit 20, the user can select two modes of "shooting mode" and "image browsing mode" as the mode, Quot; mode and the &quot; enhanced image correction mode &quot;.

The display unit 30 is a display device having an LCD (Liquid Crystal Display) or the like, and performs various displays based on the display signal output from the processing unit 10. [ In addition to the icon for activating the time information and the application, various kinds of images generated by the image processing described later are displayed on the display unit 30.

The communication unit 40 is a communication device for transmitting and receiving information used in the apparatus to and from an external information processing apparatus (e.g., a personal computer) under the control of the processing unit 10. [ As a communication method of the communication unit 40, there is a type in which a wired connection is made through a cable conforming to a predetermined communication standard, a type in which a charger called a cradle is connected through a common intermediate device, Such as a connection type, can be applied.

The image pickup section 50 is an image pickup device configured to pick up an image of an arbitrary scene and has an image pickup device (semiconductor device) such as a CCD (Charge Coupled Device) image sensor or a CMOS (Complementary MOS) . The imaging unit 50 images the light emitted from the object to be imaged on the light receiving plane of the imaging element by a lens not shown and converts the contrast of light of the image into an electric signal by photoelectric conversion. The converted electric signal is converted into a digital signal by an A / D (Analog Digital) converter (not shown) and output to the processing unit 10.

The display control unit 127 controls the display unit 30 to display on the display unit 30 a screen (a confirmation screen to be described later) showing a scene to be shot by the user based on the digitized electric signal output from the imaging unit 50. [ Then, in accordance with the user's image capturing operation (for example, the operation of tapping the image capturing button of the touch panel 21), the image capturing unit 50 is converted into a predetermined format The image data is stored in the storage section 80 as the data of the photographed image. In the present specification, an image obtained as a result of imaging by the imaging section 50 is referred to as a "captured image".

An antenna 55 for a cellular phone is an antenna for transmitting and receiving radio signals for a cellular phone with a radio base station provided by a communication service provider of the smartphone 1. [

The cellular phone radio communication circuit unit 60 is a communication circuit unit of a cellular phone constituted by an RF conversion circuit, a baseband processing circuit, and the like. The cellular phone radio communication circuit unit 60 modulates and demodulates a cellular phone radio signal, .

The clock unit 70 is an internal clock of the smartphone 1, and is constituted by, for example, a crystal oscillator and a crystal oscillator which is an oscillation circuit. The timing of the clock section 70 is output to the processing section 10 at any time.

The inertial sensor 75 is a sensor for detecting the inertia information of the smartphone 1 and is constituted by, for example, a three-axis acceleration sensor or a three-axis gyro sensor. The inertial sensor 75 detects the three- The three-axis acceleration vector and the three-axis angular velocity vector are output to the processing unit 10. [

The storage unit 80 includes a storage device such as a ROM (Read Only Memory), a flash ROM, and a RAM (Random Access Memory). The storage unit 80 stores a system program for controlling the smart phone 1, And stores various programs and data for executing various application processes.

In the present embodiment, the storage unit 80 stores an image processing program 81 that is executed as image processing (see Figs. 3 and 4) as a program. The image processing program 81 includes a focus separation determination program 811 executed as a focus separation determination processing (see FIG. 5), a depth estimation program 813 executed as a depth estimation processing (see FIG. 6) An emphasized image generating program 815 executed as a generating process (see Fig. 7) and an emphasized image correcting program 817 executed as an emphasized image correcting process (see Fig. 8) are included as subroutines. These processes will be described later in detail using a flowchart.

An image data set 83 as data, emphasis image data 85, corrected emphasis image data 87, and a penalty value table 89 are stored in the storage unit 80.

The image data set 83 is constituted by various data used by the processing section 10 to perform various image processing, and an example of its data configuration is shown in Fig. In each image data set 83, a combination of data for an image picked up by the image pickup section 50 in different scenes is stored. Each of the image data sets 83 includes a plurality of input image data 833, reduction depth map data 835 generated based on the plurality of input image data 833, and a plurality of input image data 833 And the foreground image data 837 generated based on the reduction depth map data 835 are included.

The plurality of input image data 833 are information for specifying a scene for which imaging is performed by a camera such as a focal length, a focusing distance, a diaphragm value, etc. (hereinafter, Quot;), and is data composed of an input image which is a captured image picked up by the image pickup section 50, and a side information relating to the input image. As the side information of the input image in each input image data 833, the image pickup date and time 833a, an image number 833b for uniquely identifying the input image, Data 833c are stored in association with each other.

The shooting date and time 833a is stored in the input image data 833 by the processing unit 10 based on the time counted by the clock unit 70. [

In this embodiment, the processing unit 10 sets the image number 833b in accordance with a predetermined rule, and stores it in association with the input image data 833. [

For example, in the case where the imaging section 50 performs imaging by moving the focal point inward gradually while taking an image while focusing on the first front side at the time of imaging, the processing section 10 sets the imaging date and time 833a of each input image On the basis of the rule that the focal length becomes longer as the photographing time becomes later, the image number 833b is assigned in order from the input image having the earlier photographing time in order from the first. On the contrary, when the image is picked up with the focal point positioned at the rearmost position at the time of image pickup and the focus is shifted forward, the focal distance becomes shorter as the photographing time becomes later, The image number 833b may be assigned in order starting from the first. In addition, in the case where the image name is automatically given in correspondence with the camera specification at the time of image pickup (for example, when the image file name is automatically given so that xx increases in Image xx as the focusing distance increases, The image number 833b may be allocated according to the order of the image names (Image1, Image2, Image3, ...) of the image (ascending order).

In a case where the imaging unit 50 is configured to store the camera specification data as meta data such as EXIF information in the input image at the time of imaging, the image specification unit 50 refers to the camera specification, (833b) may be allocated.

The reduced depth map data 835 is data of a reduced depth map which is a reduced depth map generated by the depth map generating unit 113 using the reduced input image. The depth map is map data in which the depth, which is information indicating which of the plurality of input images is focused on which input image, is determined pixel by pixel. In the present embodiment, a depth estimation process for estimating a depth, which is information indicating which one of a plurality of input images is focused, is performed on a target pixel, and a depth map is generated based on the result of the depth estimation do. In the present embodiment, the target pixel is described as all the pixels in the image. However, it is needless to say that not all the pixels in the image but pixels included in a part of the image may be used as target pixels.

The fore-focus image data 837 is a histogram of the image composed of the pixel values of the input image estimated to be focused in all the pixels generated by using the depth map and the plurality of input images Data.

The plurality of input images or the foreground images are coded in, for example, the JPEG format, and encoded into the image No. 831, the shooting date and time 832, the type of the smartphone, the aperture value, the focal length, It is stored together with the accompanying EXIF information. At this time, the reduced depth map data 835 can be stored in the form of being included in the EXIF information as additional data. By doing so, the reduced depth map data 835 can be stored in the storage unit 80 as one JPEG file. The EXIF information may further include information (for example, reduction rate) used for generating the reduced depth map. The image data set 83 is stored in a nonvolatile memory (e.g., flash ROM).

Instead of including the reduced depth map as additional data in the EXIF information, the reduced depth map may be encoded in the JPEG format together with at least one of the plurality of input images and the foreground image and included in the image data set 83 It is also possible to do. In this case, the EXIF information of the coded image data may include information (for example, reduction rate) used for generating the reduced depth map.

As will be described in detail later, in the present embodiment, focus separation determination processing for determining whether or not the focus of a plurality of input images is appropriately separated is performed, and a reduced depth map is generated using an input image in which the focus is appropriately separated . That is, if the focus is not appropriately separated, the determined input image is not used for generation of the reduced depth map. Therefore, it is preferable to include, in the EXIF information, information indicating which of the plurality of input images has been used to generate the reduced depth map. That is, by storing the reduced depth map and the information of the input image used for generating the reduced depth map in association with each other, the generation of the foreground image, the generation of the enhanced image, and the correction of the enhanced image It is made clear which input image should be used for processing.

The emphasis image data 85 is data of an image in which the emphasized image generating section 123 emphasizes a specific pixel portion of the foreground image according to a user operation and shades another pixel portion. The emphasized image data 85 may be stored in a volatile memory (for example, a RAM) as temporary data or may be stored in a nonvolatile memory (for example, a flash ROM) in the image data set 83 Maybe.

The corrected emphasis image data 87 is image data obtained by correcting a part of the emphasized image in accordance with a user operation. The corrected emphasis image data 87 may be stored as temporary data in a volatile memory (for example, RAM) or may be stored in a nonvolatile memory (for example, flash ROM) in the image data set 83 .

The penalty value table 89 is a table in which the penalty value, which is a kind of parameter value used when the score value calculation process is performed as the appropriate solution calculation process to be described later, is set in the depth estimation process performed by the depth estimation unit 111, And stored in a volatile memory (e.g., ROM) in advance. In the score value calculation process, a cost value (cost value) corresponding to the edge intensity of the edge image at the corresponding depth of the target pixel is calculated for the target pixel, and the cost value and the penalty value are used to calculate a predetermined energy function As a score value. The penalty value is a penalty imposed when changing an input image that is focused on a target pixel and a neighboring pixel of the target pixel. In the present embodiment, the cost value corresponds to the first kind parameter value, and the penalty value corresponds to the second kind parameter value.

3 is a diagram showing an example of the table configuration of the penalty value table 89. Fig.

In the penalty value table 89, the category 891, the first penalty value 893, and the second penalty value 895 are associated with each other.

The penalty value when the input image that is focused on the target pixel and the neighboring pixels is changed by &quot; 1 &quot; in the image number 833b is the first penalty value 893. In addition, the penalty value when the input image that is focused on the target pixel and the neighboring pixels is changed to "2 or more" in the image number 833b is the second penalty value 895. "P ₁ " is set as the first penalty value 893 and "P ₂ " is set as the second penalty value 895 in the "normal time" of the category 891. However, &quot; P ₁ &lt; P _{2 &} quot ;.

In addition, in the case of "correcting the emphasized image" of the category 891, there is a case where the drag direction and the edge direction by the user satisfy a predetermined approximate condition and the case where the drag direction by the user and the edge direction are set at a predetermined distance Dissociation) condition is satisfied, another penalty value is determined. Specifically, for the former, "P ₁ -α" is defined as the first penalty value 893 and "P ₂ -α" is defined as the second penalty value 895. With respect to the latter, as a penalty value (893) of the first is "P ₁ + α", a penalty is defined as a value 895 of each of the 2 "P ₂ + α". The value of &quot; alpha &quot; can be appropriately set, for example, &quot; alpha = 1 &quot;.

2-1. First Embodiment

2-1-1. Flow of processing

4 and 5 are flow charts showing the flow of image processing executed by the processing section 10 in accordance with the image processing program 81 stored in the storage section 80. Fig. This image processing is processing for starting execution when the camera function of the smartphone 1 is selected by the user.

In the present embodiment, as the image processing, as a "photographing mode", a plurality of photographed images obtained by changing the focal distance, for example, are used as input images to generate a depth map, a full focus image, And a process that focuses on generation and display are described and illustrated. However, as in the camera function provided in the conventional smart phone, it is naturally possible to perform imaging in the &quot; normal shooting mode &quot; in which the image of the pan focus is captured and displayed without changing the focal distance. Explanations are omitted.

First, the processing unit 10 determines the mode selected by the user through the input unit 20 (step A1). When the selected mode is the &quot; photographing mode &quot; (step A1; photographing mode), the processing unit 10 changes the focal distance for the same scene according to the shutter operation of the user, 50) (step A3). For example, it is assumed that one image data set 83 for a certain scene is acquired with one shutter operation by the user.

Thereafter, the input image reduction unit 101 performs input image reduction processing for reducing a plurality of input images by using a predetermined reduction method (step A5). In this case, for example, the image is reduced to 1/2 ⁿ (n is an integer of 1 or more) so that the image information of the original input image is lost as much as possible. The reduction of the input image can be realized by a known technique such as downsampling. The reduced input image is stored in association with the original input image to be reduced.

Thereafter, the alignment unit 103 performs alignment processing of a plurality of reduced input images stored in the storage unit 80 (step A7).

There is a high possibility that positional deviation occurs between the input images when photographing by the user is performed in a state in which the photographer is in a hand, not in a stable state such as a state where the smartphone 1 is fixed. If there is a positional shift, the position of a pixel to be output as focusing or non-focusing is shifted, and as a result, a depth map including incorrect depth information may be generated. Here, the reduced input image is aligned. For the alignment processing, various techniques can be applied. For example, a known alignment technique using an optical flow can be applied. This method includes a block matching method and a gradient method.

Next, the contrast correction section 105 performs a contrast correction process (step A9). There is a problem that the edges are shifted between the input images due to the presence of strong light in the shooting scene when the edge is detected in the edge detection processing to be described later. Here, contrast correction is performed according to a predetermined contrast correction expression for the luminance of the input image. In this case, the entire image may be subjected to contrast correction, or a part of the image may be subjected to contrast correction.

When contrast correction is applied to a part of an image, it is necessary to consider the occurrence of highlight clipping (blown out highlights). The highlight clipping means that overexposure occurs in a region where strong light is irradiated, information of a seaweed (tone) is lost, and it becomes whitish. In a highlight clipped area (hereinafter referred to as a &quot; highlight clipping area &quot;), the seaweed may be entirely dulled, so that there is a possibility that an appropriate edge may not be detected when the edge is detected.

Here, a highlight clipping area is detected using a known method, and a process of locally lowering the contrast for the highlight clipping area, preferably for a larger area including the highlight clipping area, is performed. With this processing, it is possible to prevent an edge at a wrong position from being detected at the border of the highlight clipping area in the trailing edge detection processing, and it is possible to reduce the influence of the edge deviation in the depth estimation and depth map generation described later .

The above highlight clipping can also be avoided by capturing the same image in a plurality of exposures and applying HDR to be combined with one image. Applicants have devised a technique for avoiding highlight clipping using an exposure conversion function. The details are disclosed in International Publication No. 2013/054446, wherein an exposure conversion function represented by &^quot; y = a x ^b &quot; is defined, and a minimum inversion method using sample data of pixel values of an input image of low exposure The exposure conversion function is estimated by estimating the values of the exposure conversion parameters (a, b) included in the exposure conversion function. Then, by using the obtained exposure conversion function to correct the seawater value of the target image, an object image in which the seawater dullness is eliminated is generated. By performing edge detection on the target image generated in this way, it is possible to reduce the influence of the edge deviation in depth estimation and depth map generation, which will be described later.

Next, the edge detection unit 107 performs edge detection processing on the contrast-corrected image corresponding to each of the plurality of reduced input images obtained in the contrast correction processing (step A11). In this embodiment, an edge in the contrast-corrected image is detected as a feature of the input image. Specifically, for example, as edge detection processing, edge detection is performed using a Laplace filter. However, instead of using a Laplace filter, a filter for edge detection such as a Sobel filter, a Roberts filter, a Forsen filter, or a range filter may be used. Further, the contrast-corrected image may be converted into a frequency domain by Fourier transform, and a low-frequency component may be removed to detect an edge that is a high-frequency component.

Next, the focus separation determination unit 109 performs a focus separation determination process in accordance with the focus separation determination program 811 stored in the storage unit 80 (step A13).

6 is a flowchart showing the flow of the focus separation determination processing.

First, the focus separation judgment unit 109 calculates a correlation value between edge images (step B1). The correlation value can be calculated by performing a correlation calculation for all the pixels constituting the two edge images to perform a convolution operation on all the pixels. Then, the focus separation judgment unit 109 calculates the number of pixels whose edge strength exceeds a predetermined threshold value strength for each edge image (step B3).

Thereafter, the focus separation judgment section 109 judges an edge image satisfying a predetermined focus separation condition (step B5). The focus separation condition is set such that, for example, the correlation value calculated in step B1 does not reach the first threshold value (first threshold value condition), and the number of pixels calculated in step B3 reaches the second threshold value (The second threshold condition). That is, the fact that the correlation value satisfies the predetermined low threshold value condition and the pixel number satisfies the predetermined high threshold value is set as the focus separation condition.

However, instead of the number of pixels whose edge strength exceeds a predetermined threshold value strength, the ratio of pixels whose edge strength exceeds a predetermined threshold value strength may be calculated. In this case, when the correlation value does not reach the first threshold value (the first threshold value condition) and the ratio of the pixels whose edge strength exceeds the predetermined threshold intensity reaches a predetermined threshold value (The second threshold value condition) as the focus separation condition. That is, it may be determined that the correlation value satisfies a predetermined low threshold value condition and that the ratio of pixels satisfies a predetermined high threshold value as the focus separation condition.

Next, the focus separation determination unit 109 determines the input image corresponding to the edge image determined to satisfy the focus separation condition as an input image to be used for depth estimation (step B7). Then, the focus separation determination unit 109 erases the input image that does not satisfy the focus separation condition (step B9), and ends the focus separation determination processing.

However, in the above case, it is also possible to make an error notification when there is an input image that does not satisfy the focus separation condition, and prompt the user to take the image again. That is, the focal distance may be changed to the image pickup section 50, and the image pickup of the same scene may be performed again.

Returning to Fig. 4, after performing the focus separation determination processing, the depth estimation unit 111 performs depth estimation processing according to the depth estimation program 813 stored in the storage unit 80 (step A15).

7 is a flowchart showing the flow of the depth estimation processing.

The depth estimation unit 111 performs processing of the loop A for each line constituting the reduced input image (steps C1 to C9). In the processing of the loop A, the depth estimating unit 111 executes an appropriate arithmetic processing based on the energy function on the pixels included in the line (step C3).

In an appropriate arithmetic processing, a score value arithmetic processing for calculating a score value for each of a plurality of input images with respect to pixels constituting the input image is executed. Specifically, based on the energy function, the score value of each pixel constituting the input image is calculated. The score value includes a cost value determined for the target pixel corresponding to the edge intensity of the input image of the corresponding depth of the target pixel, a first input image that is focused on the target pixel, The score value for each input image is calculated by using the penalty value determined as the penalty for the case where the second input image that is focused on the neighboring pixels is different. Then, the input image in which the calculated score value satisfies the predetermined high value condition is estimated as the focused input image.

The energy function used in the present embodiment will be described.

According to the discovery of the present inventor, the intensity (edge intensity) of an edge obtained from an edge image has a correlation with the degree of focusing at the focal length. It is considered that an input image having the strongest edge among the plurality of input images can be said to be the most focused input image in the pixel. From this logic, based on the edge information, a depth appropriate solution for each pixel is derived, and a depth map is generated as the set. For generating the depth map from the edge image, for example, the energy function shown in the following equation (1) is used.

In Equation (1), "p" and "q" are pixels (coordinate values of pixels), and "N _p " represents a predetermined range in the vicinity of the pixel p. The pixel q is a pixel included in a predetermined range N _p in the vicinity of the pixel p and is a pixel near the pixel p.

&_Quot; R _x &quot; is the depth in the pixel x and corresponds, for example, to the image number 833b of the input image on a one-to-one basis. As described above, the image number 833b is assigned in order of 1 based on a predetermined rule (for example, a rule that the focal length becomes longer or shorter as the photographing time becomes later). In this case, since the image number 833b is in ascending or descending order of the focal length, it can be said that it represents the depth of the input image.

"C (p, R _x )" is the cost value of the pixel p at the depth R _x , and this cost value is calculated as the reciprocal of the edge intensity. That is, the higher the edge strength, the smaller the cost value. T [] is a function that takes "1" when the condition in the parentheses is satisfied, and "0" when it is not satisfied.

Quot; P (p) &quot; is the penalty value in the pixel p. "P ₁ " is a first penalty value that is a penalty value when the depth is changed by one, and "P ₂ " is a second penalty value that is a penalty value when the depth is changed by two or more. The first penalty value and the second penalty value are penalty values determined in the "normal time" of the penalty value table 89. Quot; P ₁ &lt; P _{2 &} quot ;, as described above.

The energy function of the equation (1) is a function for selecting the information of the most focused input image among the plurality of input images as much as possible for the target pixel p, A suitable solution for selecting which of the plurality of input images to use the pixel value of the input image is derived for each pixel. If appropriate, the score value, which is the value of this energy function, takes the maximum value. In this energy function, the score value is determined corresponding to the pixel value (luminance value) in the edge image. When the input image having the maximum pixel value is sequentially selected, the output of the energy function becomes maximum Respectively. This is the cost value of the first term on the right side of equation (1).

On the other hand, when the input image to be selected is changed in the vicinity pixel q of the target pixel p, the penalty value is exceeded, that is, the predetermined value is subtracted. This is the penalty term of paragraphs 2 and 3 on the right side of equation (1). Specifically, when the first input image to be focused on the target pixel p is different from the second input image to be focused on the neighboring pixel q, a predetermined penalty value (a first penalty value and a second penalty value 2) is subtracted from the score value.

Further, as described above, the second penalty value is determined to be larger than the first penalty value (P ₁ <P ₂ ). This is because the greater the difference between the depth "R _q" which is estimated with respect to the depth "R _p", and the neighboring pixel q that is estimated for the target pixel p, means for subtracting a value from greater penalty score value.

From the above, the energy function of Equation (1) is configured such that suppression is applied so as not to change the input image as much as possible in neighboring pixels while selecting the input image that takes the maximum possible pixel value. By using the cost value, it is possible to easily select an input image having a strong edge with respect to each pixel as a focused image, thereby obtaining a depth map reflecting the strength of the edge. In addition, by using the penalty value, it is possible to obtain a smooth depth map in which the estimated depth is prevented from being uneven, and the depth of field is smoothly changed.

However, the input image captured by the image pickup section 50 is, in principle, another captured image obtained by changing the focal distance, but there is no problem even if an image having the same focal length is included in the input image. In this case, for the input image of the same focal distance, the penalty value may be set to &quot; 0 &quot;

In this embodiment, in order to facilitate the calculation of the score value in the above-mentioned appropriate arithmetic processing, the energy function of the equation (1) is one-dimensionalized, the score value is computed for each line constituting the input image, Calculate the depth appropriate solution. If the equation (1) is linearized, the following equation (2) is obtained.

In equation (2), &quot; L _r &quot; is a penalty function in line r. &Quot; _Lr (p, d) &quot; is the cost value in the pixel p in the line r. Further, &quot; C (p, r) &quot; is a cost value when the depth of pixel p is r.

In step C3, a score value calculating process is executed based on the energy function of equation (2), and a score value is calculated for each pixel constituting the line. Then, the depth estimation unit 111 determines a depth R _x at which the score value becomes the maximum for the pixel x for each of the pixels included in the line (step C 5). The depth estimating unit 111 then judges the input image of the image number 833b corresponding to the determined depth R _x as the focused image of the pixel x (step C7) .

If the processing of steps C3 to C7 is performed on all the lines constituting the input image, the depth estimating unit 111 ends the processing of the loop A (step C9) and terminates the depth estimation processing.

Returning to the image processing of Fig. 4 and performing the depth estimation processing, the depth map generation unit 113 performs the reduced depth map generation processing (step A17). Specifically, based on the result of the depth estimation processing, a map in which a bit value corresponding to the estimated depth is assigned to each pixel constituting the reduced input image is called a reduced depth map. However, the number of bits may be set corresponding to the number of input images used in the depth estimation processing, for the generated reduced depth map. For example, if the number of input images is 3 to 4, the depth map may be represented by 2 bits, and if the number of input images is 5 to 8, the depth map may be expressed by 3 bits. The generated reduced depth map is stored in the image data set 83 of the storage unit 80 in association with the plurality of input images.

Next, the depth map restoration unit 115 performs a reduced depth map restoration process (step A19). Specifically, the reduced depth map is enlarged to an original size using a predetermined restoration method, and this is referred to as a restoration depth map.

Next, the depth map correcting unit 117 performs a depth map correcting process (step A21). Specifically, the depth map correcting unit 117 corrects the restoration depth map by using one input image selected from a plurality of input images as a correction image. In this case, for example, among the plurality of input images, the input image whose focal distance is closest to the intermediate value can be selected as the correction image.

However, a synthesized image obtained by synthesizing two or more input images selected from a plurality of input images may be used as a correction image.

The correction of the depth map can be realized by performing weighted averaging processing using the pixel value of the correction image. This weighted average processing is given by the following equation (3).

In Equation (3), "p" is a correction target pixel and "q" is a neighboring pixel located in the vicinity of the correction target pixel. "G (p, q)" is a function that increases as the distance between the pixel p to be corrected and the neighboring pixel q becomes closer, and is given by, for example, a Gaussian function. "I (p)" is the luminance value in the pixel p to be corrected in the correction image, and "I (q)" is the luminance value in the pixel q in the vicinity of the correction image. I (p) and I (q) are functions whose values become larger as the luminance values I (p) and I (q) become closer to each other. "D (q)" is a value at a nearby pixel q in the depth map, and "D '(p)" represents a correction value at the pixel p to be corrected in the depth map obtained by weighted averaging. The correction depth map thus obtained is used as a depth map in subsequent processing.

Here, in order to reduce the processing load of the weighted averaging process of the equation (3), the explanation of the principle principle is omitted, but the weighted averaging process may be performed using the quantization code according to the following equation (4).

Next, the full-focus image generation unit 119 performs a foreground image generation process (step A23). Specifically, on the basis of the depth map corrected in step A21, a pixel value to be selected from among a plurality of input images is extracted, and a foreground image composed of pixel values of an input image having a high degree of focusing in all the pixels is generated do. The generated front focus image data 837 is stored in correspondence with the input image data 833 and the reduction depth map 835 in the image data set 83 of the storage unit 80.

5, the display control unit 127 controls the display unit 30 to display a foreground image (step A25). Thereafter, the processing unit 10 determines whether or not the user has selected the &quot; enhanced image generation mode &quot; through the input unit 20 (step A27) The unit 123 performs emphasis image generation processing in accordance with the emphasis image generation program 815 stored in the storage unit 80 (step A29).

8 is a flowchart showing the flow of the enhanced image generation processing.

The emphasized image generating section 123 judges the designated area according to the user operation (step D1). Specifically, the tapped area of the touch panel 21 is recognized, and the tapped area is determined as the designated area. Thereafter, the emphasized-image generating unit 123 selects representative pixels among the designated areas (step D3).

Since the points tapped by the user's fingers have an area, the corresponding coordinates are not limited to one but span a plurality of coordinates. As a result, a plurality of pixel values are extracted, and it is impossible to uniquely determine a focal distance (to maintain focusing) to be emphasized. Thus, in the present embodiment, one pixel is selected as a representative pixel from among the designated areas constituted by a plurality of extracted pixel values.

In this case, the representative pixels may be selected by, for example, (1) a pixel nearest to the center of the designated area, and (2) a pixel included in the designated area corresponding to the deepest or shallowest pixel in the depth map , And (3) a pixel having the closest pixel value to the average value of the pixel values of the pixels included in the designated area, as the representative pixel.

Thereafter, the enhanced image generation unit 123 selects the depth of representative pixels based on the depth map (step D5). That is, it is determined which depth of the pixel selected as the representative pixel corresponds to which depth, and the determined depth is selected.

However, instead of selecting the representative pixel and selecting the depth of the representative pixel, it is also possible to select one depth using the depth map without selecting the representative pixel. More specifically, the step D3 may be omitted, and in step D5, for example, (4) the depth of the area of the depth map which corresponds to the largest area occupied by the corresponding depth pixel may be selected.

Next, based on the depth map, the emphasized-image generating unit 123 specifies a pixel corresponding to the depth selected in step D5 (step D7). That is, based on the bit value indicating the depth of each pixel of the depth map, it is determined which pixel corresponds to the pixel of the depth determined in step D5.

Thereafter, the enhanced image generation unit 123 generates an enhanced image (step D9). Specifically, a process of shading the entire foreground image with a predetermined degree of shading is performed using a shading filter. As the shading filter in this case, for example, a known filter such as a Gaussian filter can be applied. Then, the pixel value of the pixel specified in step D7 is synthesized with the shading image obtained as a result of the shading processing, and the processing result is regarded as an enhanced image (synthesized image). Then, the display control unit 127 updates the display of the display unit 30 with the generated enhanced image (step D11).

When a pixel having a focal distance to be emphasized is combined with a shading filter applied to a fore-focus image constantly, there arises a phenomenon that an area that becomes an unnatural pixel value such as a contour near the boundary between focusing and non-focusing occurs . This phenomenon is caused by the influence of the pixel at the focal distance to be emphasized even though the pixel is not the focal distance to be emphasized but near the boundary of the object in the shooting scene. In order to avoid this phenomenon, a weighting is given to a portion which is not a pixel to be emphasized, and a shading filter is applied to a fore-focus image to perform synthesis. More specifically, shading processing is performed with respect to an image portion other than the pixel specified in Step D7, among the full-focus image, by making the degree of shading larger than the predetermined degree of shading.

However, instead of the above, the shading process may be performed with the shading degree smaller than the predetermined shading degree for the pixel specified in step D7.

Next, the emphasized-image generating unit 123 determines whether there has been a user operation continuously (step D13). If it is determined that there is a user operation subsequently (step D13; Yes), the enhanced image generating unit 123 returns the process to step D1. This includes a case where the user performs a drag operation from a position once tapped by the user, in addition to the case where the user taps a completely different position. That is, when the drag operation is performed by the user, the representative pixels are dynamically selected according to the sequential change of the designated area by the drag operation, and the depth of representative pixels is dynamically determined based on the depth map. The emphasized image generating unit 123 dynamically generates a composite image corresponding to the representative pixel based on the depth dynamically determined as the designated area is changed by the drag operation, Controls to dynamically display the composite image dynamically generated on the display unit 30 in accordance with the drag operation.

If it is determined in step D13 that there is no user operation (step D13; No), the enhanced image generation unit 123 determines whether or not the enhanced image generation mode is to be terminated according to the user operation (step D15) (Step D15; No), the process returns to Step D13. On the other hand, if it is determined that the process has ended (step D15; Yes), the process of generating the enhanced image is terminated.

5, the processing unit 10 determines whether the emphasized image is being displayed on the display unit 30 (step A31). If it is determined that the emphasized image is being displayed (step A31; Yes) Whether or not the &quot; enhanced image correction mode &quot; is selected through the input unit 20 (step A33). If it is judged that the selected image has been selected (step A33; Yes), the enhanced image correction unit 125 performs the enhanced image correction process according to the enhanced image modification program 817 stored in the storage unit 80 (step A35) .

9 is a flowchart showing the flow of the enhanced image modification processing.

The emphasized image correction unit 125 sets the area dragged by the user through the touch panel 21 as a designated area (step E1). Then, the emphasized-image correcting unit 125 selects a representative pixel in the designated area (step E3). The representative pixel can be selected by the same method as in the case of selecting representative pixels in step D3 of the enhanced image generation processing in Fig.

Next, the enhanced image correction unit 125 selects the depth of representative pixels based on the depth map (step E5). That is, it is determined which depth of the pixel selected as the representative pixel corresponds to which depth, and the determined depth is selected.

Thereafter, the emphasized image correction unit 125 calculates the cosine similarity as an index value indicating the relative relationship between the drag direction by the user and the direction of the edge included in the designated area in the input image of the depth selected at step E5 (Step E6). The cosine similarity is one of the similarity calculation methods in the vector space model, and is an index value expressing the angle formed by the vectors and the closeness. The similarity of the cosines means that the vectors are similar when they are close to "1", and the vectors are not similar when they are close to "0". However, instead of the cosine similarity, the cosine distance between the drag direction and the edge direction may be calculated.

The emphasized image correction unit 125 sets a penalty value for a pixel in the designated area based on the cosine similarity calculated in step E5. Specifically, when the cosine similarity degree exceeds a predetermined threshold value (e.g., 0.7), it is determined that the drag direction and the edge direction satisfy the approximate condition, and the penalty value set in the penalty value table 89 Quot; P _{1 -} alpha &quot; and the second penalty value &quot; P _{2 -} alpha &quot;). When the cosine similarity degree is equal to or less than the threshold value, it is determined that the drag direction and the edge direction satisfy the divergence condition, and the penalty value set in the penalty value table 89 (the first penalty value is "P ₁ + 2 &quot; is set to &quot; P ₂ + alpha &quot;).

As a result, when the drag direction and the edge direction satisfy the approximate condition (for example, when the drag direction and the edge direction are parallel), the penalty value becomes smaller as compared with the normal case, The penalty for the depth difference of the pixels of the pixel is weak. As a result, with respect to the pixels included in the designated area, the tolerance of the input image to be selected as being focused is increased, and the likelihood that the input image is selected with a wider depth width increases. Therefore, (Emphasizing) action occurs.

On the other hand, when the drag direction and the edge direction satisfy the disparity condition (for example, when the drag direction and the edge direction are orthogonal to each other), the penalty value becomes larger as compared with the normal case, The penalty for the difference in the depth of view becomes strong. As a result, with respect to the pixels included in the designated area, the tolerance for the input image to be selected as being focussed is reduced, and a strong restriction is imposed on the width of the depth for selecting the input image. Thus, (Erase) action occurs.

However, when the drag direction and the edge direction satisfy the approximate condition, the penalty value is set to be smaller than the penalty value in the case where the drag direction and the edge direction satisfy the approximate condition. It may be set to the same value as the normal time. In this case, the depth map generated at the normal time is saved, and edge information approximate to the drag direction is stored.

Next, the emphasized image correction unit 125 executes a score value calculation process based on the energy function of the equation (1) for pixels in the designated area (step E9). Then, the emphasized image correcting unit 125 judges the input image having the maximum score value for the pixels in the designated area as the focused input image (step E11).

Thereafter, the emphasized-image correcting unit 125 performs a modified-depth-map generating process (step E13). Specifically, based on the determination result in step E11, a depth map is generated in which a bit value corresponding to an input image determined to be focused on each pixel constituting the input image is assigned, and this is called a correction depth map.

Thereafter, the emphasized-image correction unit 125 performs a corrected-emphasized-image generation process (step E15). More specifically, an enhanced image obtained by modifying the pixel value of the designated area is generated using the modified depth map generated in step E13.

Next, the emphasized image correction unit 125 determines whether or not the drag operation has been continuously performed by the user (step E17). If it is determined that the drag operation has been performed (step E17; Yes), the process returns to step E1. On the other hand, if it is determined that the enhancement image correction mode has not been reached (step E17; No), the enhanced image modification unit 125 determines whether or not the user has performed an operation to end the enhanced image modification mode (step E19).

If it is determined that the operation for ending the enhanced image modification mode has not been performed (step E19; No), the enhanced image modification unit 125 returns the process to step E17. On the other hand, if it is determined that the operation for ending the enhanced image modification mode has been performed (step E19; Yes), the enhanced image modification processing is terminated.

5, if the enhanced image correction process has been performed, the processing unit 10 determines whether the user has selected a shooting mode or an image viewing mode through the input unit 20 (step A37). If it is determined that the shooting mode or the image viewing mode has been selected (Step A37; Yes), the process returns to step A1. On the other hand, if it is determined that the camera function is not selected (step A37; Yes), the processing unit 10 determines whether or not the user has performed the ending operation of the camera function through the input unit 20 (step A39).

If it is determined that the ending operation of the camera function has not been performed (step A39; No), the processing section 10 returns the processing to step A27. On the other hand, if it is determined that the ending operation of the camera function has been performed (step A39; Yes), the processing unit 10 ends the image processing.

On the other hand, the display control unit 127 is stored in the image data set 83 of the storage unit 80 (step A1) Focus image is displayed on the display unit 30 (step A41). Then, in accordance with the user's operation from the input unit 20, the processing unit 10 selects one fore-focus image among the foreground images displayed in the list (step A43).

Next, the processing section 10 extracts the reduced depth map stored in the image data set 83 in correspondence with the fore-focus image selected in step A43 (step A45). Then, the depth map restoration unit 115 performs reduction depth map restoration processing for restoring the extracted reduction depth map (step A47). The reduced depth map stored in the image data set 83 is restored in order to return it to its original size since it is not available as a depth map as it is reduced.

Here, the restoration depth map obtained by restoring the reduced depth map has a problem that the restoration depth map deviates from the edge portion in the foreground image. If the emphasized image generation process described above is performed in this state, the subject to be emphasized can not be accurately reflected, and a part of the background is included in the focusing area, or conversely, a part of the subject is made the non-focusing area . Therefore, it is inappropriate to use the reconstruction depth map obtained by reconstruction as a depth map as it is. Here, the depth map correction unit 117 performs a restoration depth map correction process for correcting the restoration depth map (step A49).

Specifically, the depth map correcting unit 117 reads a fore-focus image stored in association with the reduction depth map extracted in step A39, performs a weighted averaging process using the foreground image as a correction image, Modify the depth map. The weighted averaging process in this case can be realized by using the same equation as the above-mentioned equation (3).

In this case, in the equation (3), "p" is the pixel to be corrected and "q" is the neighboring pixel located in the vicinity of the pixel to be corrected. "G (p, q)" is a function that increases as the distance between the pixel p to be corrected and the neighboring pixel q becomes closer, and is given by, for example, a Gaussian function. "I (p)" is the luminance value in the pixel p to be corrected of the foreground image, and "I (q)" is the luminance value in the pixel q in the vicinity of the foreground image. I (p) and I (q) are functions whose values become larger as the luminance values I (p) and I (q) become closer to each other. "D (q)" is a value at a nearby pixel q in the depth map, and "D '(p)" represents a correction value at the pixel p to be corrected in the depth map obtained by weighted averaging. The correction depth map thus obtained is used as a depth map in subsequent processing.

However, instead of using the foreground image as the correction image to correct the reconstruction depth map, it is also possible to modify the reconstruction depth map by using the input image selected from the plurality of input images as the correction image. Also in this case, for example, among the plurality of input images, an input image whose focal distance is closest to the intermediate value can be selected as the correction image. Further, the reconstruction depth map may be modified using a synthesized image obtained by synthesizing two or more input images selected from a plurality of input images as a correction image.

If the restoration depth map correction processing is performed in step A49, the processing section 10 shifts the processing to step A25. That is, the display control section 127 controls the display section 30 to display the foreground image selected in step A43. The subsequent processing is the same.

2-1-2. Experiment result

Next, with respect to the above-mentioned processing, the results of actually performing experiments using images are shown.

Fig. 10 shows two images captured by changing the focal distance for the same scene.

Fig. 10 (1) is an input image focused on a puppet portion of a rabbit having a root end, and Fig. 10 (2) is an input image focused on a tree portion as a background.

The edge images obtained by performing the alignment processing and the contrast correction processing on these input images and then performing the edge detection processing are shown in Figs. 10 (3) and 10 (4), respectively. Fig. 10 (3) is an edge image obtained by performing edge detection processing on the input image of Fig. 10 (1). Since the image is focused while focusing on the doll portion of the rabbit, And is detected as an edge. On the other hand, FIG. 10 (4) is an edge image obtained by performing the edge detection processing on the input image of FIG. 10 (2) and focuses on the background part of the tree, Is detected.

Fig. 11 (1) shows a depth map generated by appropriately performing arithmetic processing using the two edge images shown in Figs. 10 (3) and (4). Since the number of input images is two, the depth map is represented by one bit. It can be seen that the dolphin part of the rhizome and the background part of the tree are properly separated.

Fig. 11 (2) shows a foreground image generated using the depth map of Fig. 11 (1) and two input images. In this case, it can be seen that a fore-focus image in a state in which both the pupil part of the rabbit and the part of the background tree are focused is generated.

FIG. 12A shows a depth map reconstructed by the depth map reconstruction process, and FIG. 12B shows a modification depth map generated by the depth map modification process. It can be seen from this that the reconstruction depth map is subjected to the weighted averaging processing using the selection input image to obtain a reconstruction depth map in which the border between the dolphin part of the rabbit and the part of the background tree is confirmed.

13 is a diagram showing a result of generating an enhanced image in the enhanced image generation processing. 13 (1) shows the result of constant shading processing on the foreground image, and FIG. 13 (2) shows the shading image of FIG. 13 (1) The results are shown in Fig.

In Fig. 13 (2), an area with an unnatural marking such as a contour appears around the doll of the rabbit. This phenomenon is caused by the influence of a pixel at a depth to be emphasized even though it is a pixel not a depth to be emphasized in the vicinity of a boundary between a doll portion of a rabbit and a part of a background tree. Fig. 13 (3) shows the result of shading processing on a foreground image by reducing the weight of the pupil portion of the rabbit, which is the root of the rabbit. 13 (4) shows the result of synthesizing the shading image of FIG. 13 (3) in order to emphasize the doll portion of the rabbit as the foreground. It can be seen from this that the unnatural display such as the outline which occurred in the vicinity of the doll of the rabbit, which can be seen in FIG. 13 (2), is eliminated.

Fig. 14 (1) shows that the emphasized image generated in the emphasized image generation processing is modeled to show only the doll portion of the foreground rabbit. The doll portion of the rabbit is tapped by the user, whereby an emphasized image in which the doll portion of the rabbit is emphasized is displayed on the display section 30. [ In this figure, it can be seen that a part of the right ear of the rabbit is inwardly deformed. This is because, in the depth map shown in Fig. 12, the depth of the right ear of the rabbit doll is not appropriately estimated, and some inaccurate depth maps are generated.

When an input image having a high focusing degree is selected by using the above-described energy function, an appropriate input image may not be selected depending on the setting of the coefficient in the energy function. As a result, The pixels other than the object portion in the distance may be displayed without being shaded.

Here, among the displayed emphasized images, the user who has viewed the pixels other than the subject portion at the object focal length without being shaded selects the emphasized image correction mode. Then, in the emphasized image modification mode, a drag operation is performed so as to follow the portion to be corrected. Specifically, in FIG. 14 (1), a portion of the region R1 in the image is dragged to follow the user. Then, the processing section 10 performs the re-calculation by changing the coefficient of the energy function (more specifically, the penalty value) for the pixel in the designated area R1 with the area R1 as the designated area R1, Correct the emphasized image by recalculation.

Specifically, the degree of similarity between the drag direction and the edge direction is calculated, and the penalty value is set corresponding to the calculated value. Here, it is determined that the cosine similarity is less than or equal to the predetermined threshold value, and when it is determined that the drag direction and the edge direction satisfy the approximate condition, the first penalty value is set to &quot; P _{1 -} Is recalculated as &quot; P _{2 -} alpha &quot;. That is, the penalty value is re-calculated by setting the penalty value lower than normal. As a result, as described above, an effect of emphasizing the edge relative to the drag direction is generated, and a depth map in which the distribution of the depths is largely modified with respect to the depth map generated at the normal time is obtained for the designated region R1 Loses. By modifying the synthesized image based on the modified depth map, the edge portion of the right ear of the deficient rabbit is emphasized and supplemented. As shown in Fig. 14 (2), in the designated region R1, A corrected emphasis image in which the shape of the right ear is appropriately modified is displayed on the display screen.

2-1-3. Action effect

According to the first embodiment, in the smartphone 1, the edge detection means 107 detects an edge as a feature from a plurality of input images, and detects the intensity of the detected edge as a feature amount. The depth estimating unit 111 uses the intensity of the edge detected by the edge detecting unit 107 to determine whether the target pixel is focused on which input image among the plurality of input images Estimate the depth of field. The depth map generator 113 generates a depth map based on the estimation result of the depth estimator 111. [

More specifically, the input image reduction unit 101 reduces the plurality of input images, respectively, and detects the edges from the reduced input image. Then, an appropriate calculation process is performed using the intensity of the detected edge to estimate the depth of the target pixel, and a depth map is generated based on the estimation result.

In the estimation of the depth, a predetermined appropriate calculation process based on the edge intensity of the edge detection image corresponding to each of the plurality of input images obtained as the detection result of the edge detection unit 107 is performed to estimate the depth. In an appropriate arithmetic process, a score value arithmetic process is performed. In this score value calculation process, when the cost value is calculated using the edge intensity for the target pixel and added to the score value, and the input image to be selected is changed with respect to the pixel near the target pixel, &Lt; / RTI &gt; In other words, suppression is applied so as to select an input image having the strongest possible edge strength for the target pixel while selecting an input image whose focus is as close as possible in the nearby pixels. It is possible to correctly estimate the depth by performing such a score value calculation process and determining an input image whose score value satisfies a predetermined high value condition as the focused image.

The depth map is information for specifying which input image is focused (i.e., depth) with respect to the pixels of each coordinate, and coordinates of the focused pixel (i.e., focused area) Quot ;. In this embodiment, a fore-focus image is generated using the depth map generated in this way and a plurality of input images, and displayed on the display unit 30. [

A fore-focus image displayed on the display unit 30 is used for generating an enhanced image in accordance with a user operation. Specifically, one depth is selected corresponding to the user's tap operation or drag operation with respect to the touch panel 21. [ Then, the pixels of the selected depth are specified based on the depth map, and an enhanced image is generated. Specifically, a shading process is performed on a foreground image with a predetermined degree of shading, and an enhanced image obtained by synthesizing the pixel value of the input image having the corresponding depth of the specified pixel with the image of the process result is generated and displayed on the display unit 30 ). This makes it possible to generate an emphasized image focused on a part designated by the user and to view the emphasized image to the user.

Further, the emphasized image is modified according to a user operation. Specifically, the depth estimating unit 111 re-estimates the depth of a pixel included in the designated area designated by the user when the user operation is detected. Then, the depth map generating unit 113 regenerates the depth map based on the result of the re-estimation of the depth, and generates a corrected high-intensity image based on the regenerated depth map. Thereby, it becomes possible to correct a portion designated by the user in the emphasized image, and to provide a corrected emphasized image corresponding to the request of the user.

2-2. Second Embodiment

In the first embodiment, a plurality of input images that have been picked up by changing the focal distance are reduced, and a depth estimation process is performed on the reduced input image to generate a reduced depth map. The second embodiment is an embodiment in which a plurality of input images are made as they are without reduction, and the same image processing as that of the first embodiment is performed on the plurality of input images to generate a depth map.

Fig. 15 is a flowchart showing a portion corresponding to Fig. 4 of the image processing in the second image processing executed by the processing section 10 instead of the image processing in Fig. 4 and Fig. However, the same reference numerals are given to the same steps as in the image processing.

First, the processing unit 10 determines the mode selected by the user through the input unit 20 (step A1). When the selected mode is the &quot; photographing mode &quot;, the processing unit 10 controls the imaging unit 50 to take an image of a plurality of images by changing the focal distance for the same scene in accordance with the shutter operation of the user (Step A3). A plurality of input images are stored in the image data set 83.

Thereafter, the contrast correction unit 105 performs the alignment processing (step A7) and the contrast correction processing (step A9) on each of the plurality of input images having different depths stored in the image data set 83. Then, the edge detection unit 107 performs edge detection processing on each of the plurality of input images corrected by the contrast correction processing (step A11).

Thereafter, the focus separation judgment unit 109 performs a focus separation judgment process on each of the edge images obtained in the edge detection process (step A13). Then, the depth estimation processing performs the depth estimation processing using the edge image determined that the focus is properly separated by the focus separation determination unit 109 (step A15).

Thereafter, the depth map generation unit 113 performs depth map generation processing (step F17). Specifically, based on the result of the depth estimation processing, a map is generated in which a bit value corresponding to a depth estimated to be focused on each pixel constituting the input image is allocated, and this is used as a depth map. However, the depth map to be generated may be set to the number of bits corresponding to the number of input images used in the depth estimation processing. For example, if the number of input images is 3 to 4, the depth map may be represented by 2 bits, and if the number of input images is 5 to 8, the depth map may be expressed by 3 bits.

Thereafter, the forehead image generation unit 119 performs a forehead image generation process (step F19). Specifically, on the basis of the depth map obtained in the depth map generation processing, a pixel value to be selected from a plurality of input images is extracted, and a fore-focus image composed of pixel values of an input image having a high degree of focusing . The generated fore-focus image is stored in association with a plurality of input images in the image data set 83 of the storage unit 80. [

Thereafter, the depth map reducing section 121 performs a depth map reducing process (step F21). Specifically, the depth map generated in step F17 is reduced using a predetermined reduction method. In this case, for example, when the depth map is restored to its original size, the image is reduced to 1/2 ⁿ (n is an integer of 1 or more) so that the image information of the original depth map is not lost as much as possible. The reduction of the input image can be realized by a known technique such as downsampling. The generated reduced depth map is stored in the image data set 83 of the storage unit 80 in association with a plurality of input images and a foreground image.

3. Third Embodiment

The third embodiment is an embodiment in which a depth map is generated by a technique different from the technique described in the first and second embodiments. In the present embodiment, the position of the camera is changed so that a parallax is generated, and a depth map is generated using a plurality of shot images picked up at different viewpoints (viewpoints). This time lag occurs. And a method of changing the position of the camera so that the parallax is generated to pick up an image at another point of time will be referred to as &quot; parallax system &quot;.

3-1. Flow of creation of depth map

Fig. 19 is a flowchart showing a portion corresponding to Fig. 4 among the third image processing executed by the processing section 10 in place of the image processing of Figs. 4 and 5 in the third embodiment. However, the same steps as those of the image processing are denoted by the same reference numerals, and repeated description thereof will be omitted.

If it is determined in step A1 that the mode is the photographing mode (step A1: photographing mode), the processing section 10 changes the position of the camera so as to control the image sensing section 50 to capture a plurality of images at different points of time (Step J3). In the method of this embodiment, in order to generate the depth map, it is necessary that a parallax is generated in the photographed image of the image of the target object. Here, it is preferable that the position of the camera is changed with respect to the user so that the user is prompted to take a desired object from a plurality of viewpoints.

Next, the processing section 10 inputs the first picked-up image as a reference input image (step J5). Then, the processing section 10 performs a feature point detecting process for detecting a plurality of feature points with respect to the reference input image (step J7). As a method of detecting a feature point, various methods can be applied. For example, a method using a corner detection algorithm of Harris, Stephens, Plessey or the like, or an affine transformation on an input image to detect a SIFT feature A known method such as a method can be applied.

Fig. 20 (1) is a diagram showing the result of detection of a feature point with respect to a reference input image. Fig.

This reference input image shows a feature point detected by performing a feature point detection process on a box having a subject on a table and a plurality of rectangular markers that are images taken by the camera at the front viewpoint and captured in the image have. It can be seen that a plurality of feature points are detected in the reference input image.

Next, the processing section 10 inputs the next captured image as an input image (step J9). Then, the processing section 10 performs minutiae tracking processing (step J11). For tracking of feature points, for example, a technique using multiple resolution analysis can be applied. More specifically, the feature point tracking process is performed in the same resolution sequentially from the coarse resolution, for example, by using the multi-resolution image created for the reference input image and the multi-resolution image created for the input image. Then, by tracing the amount of movement of the feature points in units of blocks obtained by dividing the input image into a plurality of blocks, the feature points are tracked more stably. At this time, the movement direction and the movement amount of the minutiae are calculated in which direction the respective minutiae detected in the reference input image have moved in which direction in the input image.

FIG. 20 (2) is a diagram showing the result of detection of a minutiae point on the input image.

This input image is a picked-up image obtained by picking up an image of a subject placed on a table at the left (left) point, as in (1) of Fig. The result of tracking the minutiae with respect to this input image is (3) in Fig. In (3) of Fig. 20, the starting point and the end point of each line represent the positions of the minutiae points in each image, the line color represents the moving directions of the minutiae points, and the line length represents the moving amount of each minutiae point. In this figure, it can be seen that each corresponding feature point is moving with respect to the captured image picked up at another point in time.

Next, the processing unit 10 determines whether or not to extract the input image as a starting point input image (key frame) (step J13). The starting point input image is an image that succeeds in tracing the characteristic points. For example, when the total amount of movement of the characteristic points is equal to or larger than a first threshold value (or a first threshold value) Threshold value is exceeded), the input image is extracted as the initial input image.

(Step J13; Yes), the processing section 10 stores the coordinate values of the minutiae points in the starting point input image in the storage section 80 (step J15). If it is determined that the input image is not extracted as the initial input image (Step J13; No), the processing unit 10 proceeds to Step J17.

Next, the processing section 10 determines whether extraction of the starting point input image is to be terminated (step J17). For example, in step J13, when the total amount of movement of the minutiae becomes equal to or larger than a second threshold value (or a second threshold value) that is a threshold value larger than the first threshold value, extraction of the starting point input image ends . That is, the extraction of the starting point input image ends at the point (time point) at which the overall movement amount of the minutiae point is increased to some extent. If it is determined that the extraction of the starting point input image is not to be terminated (step J17; No), the processing section 10 returns the processing to step J9. Thereby, a plurality of input images are extracted as the starting point input image.

(Step J17; Yes), the processing unit 10 uses the coordinate values of the minutiae points stored in the storage unit 80 in the step J15 with respect to the starting point input image, Fundamental Matrix) F (Step J19). The basic matrix F is a transformation matrix that transforms a basic matrix (Essential Matrix) E defining a relationship between normalized image coordinates of projection points in two images based on an epipolar geometry into an image coordinate system is a matrix of similarity transformation. However, since the basic matrix F and the basic matrix E are known, the description of the elements of the matrix is omitted.

The basic matrix F is represented by a matrix of 3 × 3, and the number of elements is nine. By setting the frobenius norm of the base matrix F to "1", the number of substantial unknowns becomes eight. Therefore, if the coordinate values of the corresponding points of the minutiae of the lowest 8 sets are obtained, the base matrix F can be obtained. Here, for the combination of the starting point input image, the base matrix F is estimated by using the coordinate values of the corresponding points of the eight or more sets of feature points stored in the storage unit 80 in step J15. Then, by using an algorithm of RANSAC (Random Sampling Consensus), for example, an appropriate basis matrix F is estimated while excluding the dropout values. Further, after RANSAC is applied, all of the minutiae within a certain error are taken as sample data, and a more appropriate basis matrix F is estimated using the minimum multiplying method.

Next, the processing section 10 converts the basic matrix F estimated in step J19 into a basic matrix E (step J21). The basic matrix E can be obtained by operating an internal parameter of the camera with respect to the basic matrix F. [ However, in the present embodiment, the internal parameters of the camera are already known.

Thereafter, the processing section 10 calculates the camera matrix using the basic matrix E obtained in step J21 (step J23). The camera matrix is a matrix indicating the relative positional relationship of the starting point (viewpoint) of the starting point input image, and is a matrix including a rotation matrix and a translational motion vector indicating camera movement. The calculation of the camera matrix can be said to be the calculation of the relative positional relationship of the starting point of the starting point input image.

Next, the processing section 10 performs bundle adjustment processing (step J25). Bundle adjustment is a technique for estimating a geometrical three-dimensional model parameter from an input image. In the bundle adjustment processing, the three-dimensional coordinates of each of the minutiae points are calculated based on the correspondence relationship between the minutiae points between the input images. Then, the three-dimensional coordinates of each calculated feature point are re-projected onto the image plane, and the re-projection error calculated as the distance between the re-projected point and the feature point is repeatedly re-estimated, . In the bundle adjustment processing, the parameter values constituting the camera matrix calculated in step J23 are estimated by performing the same processing as described above to adjust the parameter values constituting the camera matrix.

21 is a sample image when estimating the three-dimensional coordinates of the minutiae and viewing the same set of minutiae at different time points.

21 (1) shows a sample image when a group of minutiae are viewed at a frontal view, and FIG. 21 (2) shows a sample image when a group of minutiae are viewed at an upper ) Represents a sample image when the group of feature points is viewed at the left (left) view. From these sample images, it can be seen that the three-dimensional coordinates of the minutiae are correctly estimated, and that the minutiae point groups exhibit different distributions corresponding to the respective time points.

Thereafter, the processing section 10 generates a depth map using the result of the bundle adjustment processing in step J25 (step J27). In order to generate the depth map, it is necessary to calculate the position coordinates of the pixel in the three-dimensional space for each pixel. Here, a position (hereinafter, referred to as &quot; target pixel position &quot;) of a target pixel (hereinafter, referred to as a "target pixel" (Hereinafter, referred to as &quot; projection pixel position &quot;) and the camera position at the time of imaging the starting point input image.

First, the camera position at which each initial input image is captured is estimated using the camera matrix in which the parameter value is adjusted in the bundle adjustment process and the initial position of the camera stored in advance in the storage unit 80. The initial position may be an arbitrary position, and a predetermined three-dimensional coordinate value is stored in the storage section 80 as the coordinate value of the initial position. Using this initial position and the camera matrix (rotation matrix and translation movement vector), the camera position at which each initial input image is captured is calculated with reference to the initial position.

However, it is possible to store the predetermined position in the storage unit 80 in advance, but the positioning calculation is performed by using a positioning (positioning) signal such as GPS (Global Positioning System) The calculated position may be used as the initial position to perform the above calculation.

Further, for the combination of the starting point input image, a corresponding relationship is calculated on a pixel basis. More specifically, using the basic matrix E, the corresponding points are searched for only one line on the image between the starting point input images. More specifically, the position of the corresponding point of the target pixel, which is basically close to the color and close to the corresponding relationship of the neighboring pixels, is calculated using the method of semi-global optimization do.

Thus, the position of the corresponding point of the target pixel calculated with respect to the combination of the starting point input image is set as the projection pixel position. Then, for the combination of the starting point input image, the target pixel position is estimated using the two calculated camera positions and the two projected pixel positions thus calculated, and the distance from the starting point to the target pixel position is calculated. The above processing is performed with all the pixels as target pixels, and a depth map is generated in which the distance calculated for each pixel is a depth.

31 is an explanatory diagram of the principle of generation of a depth map.

31, the positions of the cameras at the time of picking up the respective initial input images with respect to the combination of the initial input image A and the initial input image B are indicated by black circles with the camera position A and the camera position B, respectively. A position where the target pixel position is projected on the initial input image A is referred to as a projection pixel position A and a position where the target pixel position is projected on the initial input image B is referred to as a projection pixel position B And are represented by white circles, respectively.

31, an intersection between a straight line connecting the camera position A and the projected pixel position A and a straight line connecting the camera position B and the projected pixel position B is estimated as the target pixel position. When the target pixel position is obtained, the distance from the viewpoint to the target pixel position can be calculated, so that the depth of the target pixel is obtained. The target pixel position is calculated using all the pixels as target pixels as described above, the distance from the viewpoint to the target pixel position is calculated, and a depth map is generated with the depth.

The following processing after generating the depth map in the above-described manner is the same as the first embodiment and the second embodiment described above.

3-2. Show Guide

In the depth map generation method described above, it is necessary for the user to move the camera to capture an image at a plurality of viewpoints in order to obtain parallax. At this time, there is a problem that the user simply does not know how to operate the camera by simply suggesting the user to move the camera, which is inconvenient. For example, if the camera is only rotated at one place, no parallax occurs and distance information can not be obtained, so it is necessary to urge the user not to perform such an operation. Here, the following processing may be performed.

After capturing the first photographed image, the processing section 10 specifies the photographed subject by, for example, tapping the portion of the camera on which the intended subject appears on the display screen of the camera. Then, the processing section 10 detects a person by performing person detection or object detection, and indicates the detected subject area. For example, a figure (hereinafter referred to as &quot; subject designation figure &quot;) slightly larger than the subject area including the subject area is displayed on the first shot image. For example, a figure of a quadrangle of a dotted line or a broken line may be displayed as a subject designation figure.

Next, the display control unit 127 determines whether the user should move the direction of the magnetic device (camera) (rightward, leftward, upward, or downward) or the distance the user moves the magnetic device A guide display for guiding the magnetic device to some extent). Concretely, a guide arrow (hereinafter, referred to as a "guide arrow") indicating a direction in which the magnetic device is to be moved in a certain direction, a figure for guiding a user to a desired photographing position Graphic form &quot;) is displayed on the display unit 30 as a guide display. For example, the guide figure has a shape and size common to the confirmation screen, and the user can grasp that it is necessary to perform an operation of moving the camera so that the preview screen and the guide figure coincide with each other.

Here, the confirmation screen is a screen that appears on the display unit 30 based on a signal output from the imaging unit 50 (imaging device) when the camera is activated in the imaging mode, and the image of the confirmation screen is output from the imaging device Signal. That is, on the confirmation screen, an image to be captured is displayed and updated in real time. The photographer can check the imaging state of the subject by looking at the confirmation screen, check whether the object of interest is displayed when shooting, or determine from which position or angle the shooting is to be performed.

Here, the direction in which the subject is to be moved may be a direction for generating parallax, and may be, for example, either vertical or horizontal. However, in the case of the vertical direction, it is generally preferable to make the horizontal direction because it is difficult to appropriately generate the parallax when the distance to the subject is long (the focal length is long). If the focal length is greater than or equal to a predetermined distance, the user can select whether to move the camera in the vertical direction or the horizontal direction. If the focal length is less than the predetermined distance, the camera may be guided to move in the horizontal direction good. Since the detection of the horizontal direction in the processing section 10 can be performed based on the detection of the gravity direction by the inertial sensor 75, the horizontal direction can be detected without depending on the photographing angle of the camera (display direction of the confirmation screen) As shown in Fig.

After storing the first shot image, the processing unit 10 tracks the detected subject area to follow the subject designation figure in the subject area, and changes the display of the guide display in accordance with the movement of the magnetic device. Here, corresponding to the movement of the magnetic device, the moving distance and the moving direction of the magnetic device from the original photographing position are changed. Here, as a moving amount of the magnetic device, for example, the moving distance and the moving direction of the magnetic device are detected, and the display of the guide display of the display section 30 is changed based on the detection result.

The inertial sensor 75 includes a three-axis acceleration sensor and a three-axis gyro sensor, and three-axis acceleration vectors and three-axis angular velocity vectors in the local coordinate system based on the sensor are output to the processing unit 10. The processing unit 10 calculates the posture of the camera by integrating the three-axis angular velocity vectors output from the inertial sensor 75 and calculates the three-axis acceleration vector in the local coordinate system based on the calculated posture, Axis acceleration vector in the absolute coordinate system. Then, the processing unit 10 integrates the three-axis acceleration vectors subjected to coordinate conversion to calculate three-axis velocity vectors. The processing section 10 performs this processing at predetermined time intervals, and calculates a moving distance and a moving direction of the magnetic device at each time interval based on the three-axis velocity vectors calculated in the time interval.

The processing section 10 determines in which direction the guide figure on the display screen is to be moved (the direction of movement of the guide figure) based on the calculated moving direction of the camera. Further, the processing section 10 determines how to move the guide figure on the display screen (movement amount of the guide figure) based on the calculated movement distance of the camera. Then, the display of the guide figure is changed based on the determined movement direction and movement amount.

Here, when a specific feature point on a subject is targeted, a travel distance required to cause a constant parallax to the feature point becomes longer corresponding to a distance (for example, a focal distance) to the subject. Therefore, the update condition of the guide figure may be different according to the distance to the subject. For example, the movement width A on the screen of the guide figure based on the movement of the constant distance X when the focal length is equal to or larger than the predetermined distance and the movement width A on the screen based on the movement of the constant distance X when the focal distance is less than the predetermined distance The moving width of the guide figure may be made different so that A < B in relation to the moving width B on the screen of the figure.

22 is a diagram showing an example of a guide display in this case.

Fig. 22 shows a display screen of the smartphone 1, which is an example of an electronic apparatus having an image processing apparatus. The entire area surrounded by the outer frame represents the display screen of the smartphone 1, and the aforementioned confirmation screen is displayed at the center of the display screen.

22 (1) shows a picked-up image in which a subject is picked up at the front view. When the image pickup by the image pickup section 50 is completed and the first shot image is stored, the confirmation screen is displayed again. When the user taps the display portion of the eye as the subject on the confirmation screen, the subject designation figure of the dotted quadrangle surrounded by the two eyes is displayed as shown in (1) of FIG.

22 (2) shows a state in which a rectangular figure indicating that the camera is moved in the horizontal direction (here, rightward direction) is displayed as the guide display. In this figure, a guide arrow in the right direction indicating that the user moves the camera in the right direction is displayed. Further, a rectangular guide figure of the same size as the confirmation screen is displayed in the form of protruding out of the right border of the confirmation screen. The user moves with the camera in the right direction, taking care that the subject is kept at the center of the confirmation screen, in accordance with the guide arrow and the guide figure.

Then, the movement direction and the movement amount of the guide display are determined by the processing section 10 as described above, and the movement of the smartphone 1 is performed in the direction opposite to the moving direction of the smartphone 1 The display position of the guide figure is changed by the amount of movement corresponding to the distance. More specifically, as shown in (3) of FIG. 22, as the user moves in the right direction, the guide figure in the display moves to the left side of the screen.

Finally, as shown in (4) of FIG. 22, when the guide figure coincides with the confirmation screen by the movement of the user (when it is judged that the outline of the guide image matches the outline of the confirmation screen) The image pickup section 50 automatically picks up the image at that position and acquires a second picked-up image having a time difference in relation to the first picked-up image.

However, the acceleration sensor and the gyro sensor included in the inertial sensor of Fig. 1 may be an inertial measurement unit (IMU) in which an acceleration sensor and a gyro sensor are packaged.

In the above description, the movement amount of the magnetic device is detected based on the acceleration vector or the angular velocity vector detected by the inertial sensor or the inertial measurement unit. Alternatively, a positioning signal transmitted from a positioning satellite, such as a GPS unit, A unit for carrying out a positioning calculation may be incorporated in an image processing apparatus and the moving amount may be detected based on the detection result of the sensor. In GPS, positioning calculation (pseudorange positioning, Doppler positioning) using a pseudo-range or Doppler frequency is performed based on a GPS satellite signal transmitted from a GPS satellite, and the position of the magnetic device can be calculated. The moving speed and the moving direction of the magnetic device may be detected based on the difference of the positioning position obtained by the positioning calculation. In the GPS, a velocity vector calculation using the Doppler frequency may be performed to calculate the velocity vector of the magnetic device. From the velocity vector thus calculated, the moving distance and the moving direction of the magnetic device may be detected.

In place of the inertial sensor 75, an inertial navigation system (INS) (Inertial Navigation System) that performs an inertial navigation operation and detects the moving distance and the moving direction independently is provided in the image processing apparatus, The moving distance and the moving direction of the image processing apparatus may be detected.

In addition to detecting the movement amount (movement direction or movement distance) of the magnetic device and changing the display of the guide display based on the detection result as described above, for example, the third image processing The display of the guide display may be changed on the basis of the result of tracking the feature point in the feature point tracking process in step J11 of Fig. That is, as a result of tracing the feature points in the feature point tracking process, it is judged as to which direction the feature point has moved (the direction of movement and distance of movement of the feature points) as a result of the tracing of the feature points, (Movement direction and amount of movement of the guide figure) may be determined, and the display of the guide figure may be changed.

4. Fourth Embodiment

4-1. Setting the default focusing position by user operation

In the above embodiment, in the emphasized image generation processing executed in the emphasized image generation mode after the foreground image is generated, the user designates an area to be emphasized in the foreground image, and a pixel included in the designated area And a focused emphasized image is generated. However, in this method, there is a problem in that, in order for the user to generate a desired image, it is necessary for the user to designate the area to be emphasized each time after the image pickup by the image pickup section 50, which is inconvenient.

Here, before the image pickup by the image pickup section 50, the user may specify a default focusing position (hereinafter referred to as &quot; default focusing position &quot;). Specifically, it is possible to allow the user to designate a default focusing position by, for example, performing a tapping operation for tapping the display screen on the confirmation screen.

23 is a flowchart in which the processing section 10 takes a portion corresponding to Fig. 4 of the image processing in the fourth image processing executed in place of the image processing in Figs. 4 and 5 in this case. The same steps as those of the image processing are denoted by the same reference numerals, and repetitive description will be omitted.

If the processing unit 10 determines that the mode is the shooting mode (step A1), the processing unit 10 determines whether or not the tap operation has been detected (step G3). If it is determined that the tap operation is detected (step G3; Yes), the processing unit 10 determines representative pixels based on the designation area (step G5). The designation area is a region tapped by the user, and the processing section 10 selects one of the designation areas as a representative pixel, as in the above embodiment. As a representative pixel selection method in this case, for example, a pixel closest to the center position of the designated area or a pixel having the closest pixel value to the average value of the pixel values of the pixels included in the designated area can be selected as the representative pixel . Then, the processing section 10 stores the determined representative pixel in the storage section 80 (step G7).

Next, the processing section 10 judges whether or not the imaging condition is established (step G9). More specifically, (1) the tap operation is detected in step G3, (2) the tapping operation of the photographing button is detected after the tap operation in step G3 is detected, and (3) (For example, 5 seconds) elapsed without detaching the finger from the tap position (detection of a long pressing operation), (4) detection of the tap operation in step G3, (Detection of a double-tap operation) is detected as an imaging condition, and when any one of these imaging conditions is satisfied, the imaging unit 50 And executes imaging.

If it is determined that the imaging condition is not satisfied (step G9; No), the processing unit 10 returns the processing to step G3. If it is determined that the imaging conditions have been satisfied (step G9; Yes), the processing section 10 performs imaging processing for changing the focal length of the imaging section 50 to pick up a plurality of images (step G11). Then, the processing section 10 shifts the processing to step A5.

Fig. 24 is a flowchart showing an example of the flow of the third enhanced image generation processing executed by the processing section 10 in place of the enhanced image generation processing of Fig. 8 in the fourth image processing. However, the same steps as those of the emphasized image generating process are denoted by the same reference numerals and repeated explanation is omitted.

The processing section 10 reads the representative pixel stored in step G7 from the storage section 80 (step J1). Then, the processing unit 10 determines the depth of the representative pixel read based on the depth map stored in the storage unit 80 (step J3). Then, the processing section 10 shifts the processing to step D7.

However, in the fourth image processing described above, when the user designates the default focusing position, it may be possible to designate the focusing position in units of pixels using a cursor or the like. In addition, the user may designate a focusing position in units of a subject by performing an operation to surround the entire subject with a finger.

Instead of performing the process of determining representative pixels in step G5, for example, a designation area is first determined according to a user operation (for example, a tap operation on a touch panel), and then an image is captured by the imaging section 50 (And corresponding depth) included in the designation area which has been initially determined after the image pickup, and generates an emphasized image in which focusing is performed on a specific pixel with respect to the depth of field . That is, the depth to be focused may be determined from the designated area designated by the user without determining the representative pixel.

In the imaging process in step G11, as in the imaging in step J3 of the third image processing described in the third embodiment, the position of the camera is changed to the user so that the target object And the image pickup section 50 picks up the image.

4-2. Setting the default focusing position by face detection / portrait detection

In the above processing, when setting the default focusing position, a person who is photographed as a subject or a face of the person is detected on the confirmation screen, and the detected person or face is set as a default focusing position, So as to generate an emphasized image. It is also possible to detect a person who is photographed as an object or a face of the person on the input image display screen on which an input image is displayed after the image is picked up and sets the detected person or face as the default focusing position So as to generate an emphasized image.

5. Fifth Embodiment

5-1. Specification of area by graphic

In the above embodiment, if a subject is aimed at a target and there is another subject at the same depth distance as that of the subject, an image with a focused focus is also generated with respect to the subject. Therefore, when a mixed image of a subject unintended by the user is generated at a depth distance equal to that of a target subject, it is impossible to shade the subject photographed by unintentionally mixing the subject by focusing the subject only on the target there is a problem.

Here, for example, a circular, elliptical, or quadrangular figure (hereinafter, referred to as &quot; area designing figure &quot;) centered on a point tapped by the user is displayed on the display screen in which the input image after the image capture is preview- By dynamically changing the size of the area designating area, it is possible to allow the user to specify an area which is not desired to be shaded by the shading process.

25 is a diagram showing an example of a figure designating area in this case.

Fig. 25 shows, as an input image, an image in which a mug, which is the object of interest, is displayed at the center. In this input image, a part of the paper cup, which is a subject unintended by the user, is displayed at the right border of the image. The user specifies the image portion of the mug by manipulating the area designating figure to emphasize the image portion of the mug which is the object of interest and to shade the image portion of the paper cup as the unintended subject.

More specifically, as shown in (1) in FIG. 25, when the user detects that the image portion of the mug, which is the object of interest, has been tapped, the display control unit 127 displays, for example, . The size of the area designating figure to be displayed in this case may be a predetermined size, and (1) in FIG. 25 shows a state in which a region designating figure of a relatively small size is displayed at the center of the mug. The area designating figure allows the size to be changed by a pinch operation using two fingers, and the size of the area designating area is enlarged by the pinch-out operation, and the size of the area designating area is increased .

As shown in (2) of FIG. 25, when the user performs pinch-out operation with two fingers on the area designating figure, the size of the area designating figure is enlarged. Then, the user adjusts the size and shape of the area designating figure by pinching-out operation so as to cover the entire mug by the area designating figure. Then, the user performs a confirmation operation, such as pressing a confirmation button, and confirms the highlighted area.

In this case, the processing unit 10, after detecting the confirmation operation in the enhanced image generation processing, sets the area surrounded by the area designation figure as the designation area. Then, for the specified area, for example, the depth to which the largest number of pixels belongs among the depths of the pixels included in the specified area is specified, and the specific depth (or the specific depth and the depth within a predetermined range ), And shading processing is performed on the other pixels to perform non-focusing. In addition, shading processing is similarly performed on an area not included in the designation area to perform non-focusing. In the shading process in this case, for example, when the subject is mixed in the area where the shading process is performed, the depth of the depth map is rewritten so that the depth distance between the subject and the background area around the subject is almost the same, Shading processing may be performed based on the written depth map.

However, the degree of shading may be changed in accordance with the difference in depth from the depth of the focusing area or the distance from the position tapped by the user, with respect to the area subjected to the shading processing. For example, among the areas not included in the designated area, the shading processing may be performed with a higher degree of shading in an area having a larger depth difference from the focusing area. Further, among the areas not included in the designated area, the shading processing may be performed with a higher degree of shading in the area where the distance from the tab position is larger. In this case, shading processing may be performed with the same degree of shading for the area having the same depth from the focusing area or the area having the same distance from the center position of the tab position.

The operation method of the area designating figure is not limited to the above. For example, the region designating figure is displayed in the first tap operation, and the operation of tapping the displayed area designating figure is repeated, The size of the used figure may be enlarged. Further, the size of the area designation figure may be enlarged as the time is lengthened, based on the time that the tab operation is continuously performed without releasing the finger after performing the tap operation.

Further, in addition to enlarging / reducing the size of the area designating figure by a pinching operation (pinch-out operation / pinch-in operation) by the user, for example, the outer edge portion of the figure for designation by the user is tapped, It may be possible to change the shape of the area designation figure by performing the operation of widening the edge to the outside or narrowing it to the inside.

The degree of shading in the shading processing may be changed corresponding to the distance or area from the center of the area designating figure. More specifically, the greater the distance from the center of the area designating figure, or the larger the area of the area designating area, the greater the degree of shading.

In addition, as described above, the area to be emphasized is not specified by using the area designating figure but the operation of filling the desired area by the drag operation and drop operation And specify an area which is not desired to be shaded.

5-2. Change the degree of shading by user operation

In addition, in the case where shading processing is performed on an area other than the area to be emphasized as described above, the degree of shading of the area to be subjected to the shading may be adjustable by a user operation. Specifically, for example, the degree of shading in the shading process may be changed corresponding to the pinch operation on the display screen, and the shading process may be performed.

More specifically, for example, on a display screen displaying a foreground image, an area to be emphatically displayed to the user is designated by a pinch operation. In this case, for example, the center position of the two fingers contacted by the pinch operation is calculated, and an area including the calculated center position pixel and pixels having the same depth on the depth map is set as the focusing area, The shading processing may be performed on an area other than the target area as a shading target area to be subjected to the shading processing.

In this case, the degree of shading in the shading process is changed corresponding to the user's pinching operation, and the shading process is performed on the shading area. Specifically, when the pinch-out operation is performed, the degree of shading is increased stepwise, and shading processing is performed on the shading target area. When the pinch-in operation is performed, the degree of shading is gradually lowered in response to the pinch-in operation, and the shading processing is performed on the shading target area.

26 is a diagram showing an example of shading processing in this case.

Fig. 26 shows a fore-focus image in which a plurality of subjects including a mini-car and a doll are photographed. In this image, it is assumed that the mini-car is set as a target object and other objects or backgrounds are shaded. 26 (1), when the user taps the image area of the minicar and performs a pinch-out operation with two fingers, the processing section 10 determines the interval between two fingers (two fingers The shading degree is stepped up in accordance with the widening of the coordinates of the two points that are in contact with each other, and shading processing is performed on image areas other than the image portion of the minicar. As a result, as shown in Fig. 26 (2), the image area of the minicar is clearly displayed and an image in which other image areas are shaded is generated.

On the other hand, in (2) of Fig. 26, when the user performs pinch-in operation with two fingers on the image area of the mini-car, the processing unit 10 determines the distance between the two fingers Shading degree is gradually lowered in response to the narrowing of the coordinate distance of the point, and the shading process is performed on the image area other than the minicar. As a result, the display state of FIG. 26 (1) is gradually returned.

Further, in the case of changing the degree of shading, the degree of shading may be set individually for each area made up of pixels at the same depth on the depth map, and shading processing may be performed. Specifically, the depth of the center position of the two fingers touching the screen in the pinch operation is set as the reference depth, the shading processing is performed on the area made up of the pixels at the reference depth with a predetermined reference shading degree, The shading processing is not performed on the area. The region may be set in the descending order of depth from the reference depth, and the degree of shading may be increased stepwise from the reference shading degree as the area is made up of pixels at depths away from the reference depth.

For example, in the case of increasing the degree of shading by the pinch-out operation, the regions consisting of pixels having the same depth are defined as regions A to D in the order of closest focal distance. In this case, when the pixel at the center position of the two fingers touching the screen in the pinch operation is included in the area B, the degree of shading is set to 20% for the area B and the degree of shading is set for the area A and the area C. 50%, the shading degree for the area D is set to 80%, and shading processing is performed with respect to all the pixels included in the area for each area, to the degree of corresponding shading.

In other words, it can be said that the effect of weakening the shading degree by the pinch-out operation is also given. For example, when the degree of shading of each area is set to 100% as a default, the degree of shading is largely reduced by setting the shading degree from 100% to 20% for the area B in the above case, and for the area A and the area C This is because the degree of shading is reduced to half by reducing the degree of shading from 100% to 50%, and by reducing the degree of shading from 100% to 80% for the region D, the degree of shading is slightly weakened. However, in this case, the area D may be set to 100% without lowering the degree of shading, so that the contrast may appear in the generated emphasized image.

However, as described above, when the pinching operation is performed in an area where the object of interest is displayed, the shading processing is not limited to the shading processing in which the degree of shading is lowered stepwise, The degree of shading may be lowered stepwise to perform shading processing. Further, in a case where a pinch-out operation is performed instead of a pinch-in operation in an arbitrary area in the image, the shading process may be performed with the degree of shading stepped down.

6. Sixth Embodiment

A focal length slider for specifying the focal length range at the time of generation of the output image or after the generation of the output image in order to freely generate an image as focusing for an arbitrary focal length range is referred to as an output image or It may be displayed outside the output image so that the user can specify an arbitrary focal length range. However, in this embodiment, the focal length and the depth correspond to each other on a one-to-one basis.

27 is a diagram showing an example of a focal length slider.

Fig. 27 shows an example of an output image in which a doll of a rabbit is displayed as a foreground at the center and a tree or a building as a background. A focal length slider S1 indicated by a bar in the lateral direction is displayed at the lower left corner of the output image. The focal length slider S1 means that the focal distance is closer to the left and focal distance is farther toward the right.

The focal length slider S1 is provided with an icon for specifying the lower limit of the right triangular shape for the user to designate the lower limit of the focal length and a icon for specifying the upper limit of the left triangular shape for the user to designate the upper limit of the focal distance, Are displayed on the bars of the distance slider S1, and the user can designate an arbitrary focal distance range by moving the icons to the left and right by the drag operation.

When the focal length range is changed, the processing unit 10 refers to the depth map stored in the storage unit 80 and specifies a pixel at a depth corresponding to the specified focal length range. Then, focusing is performed for a specific pixel, and non-focusing is performed for other pixels. However, for a pixel to be non-focused, it is preferable that the shading process is performed by increasing the degree of shading as the focal length of the pixel is away from the specified focal length range. By doing so, with respect to an area made up of pixels not included in the designated focal length range, the degree of shading becomes stronger as the focal length becomes farther away from the specified focal length range, and a natural image free from discomfort is generated.

As shown in Fig. 27 (1), when the focal length range of the focal distance slider S1 is changed to the near side (front side), the dolls of the foreground rabbit are emphasized, A shaded image is generated. On the other hand, when the focal length range of the focal length slider S1 is changed to the far side (background side) as shown in Fig. 27 (2), the background tree or building is emphasized, An image in which the doll is shaded is generated.

When the user's tap operation is detected, a depth corresponding to the focal length of the pixel included in the predetermined range near the tap position is determined based on the depth map, and the determined depth is set as the focal length slider S1. May be reflected in the range of the focal length of the lens.

For example, as shown in FIG. 27 (3), when a part of the background tree is tapped by the user, a pixel included in a circle having a predetermined radius centering on the tab position is set as a neighborhood range pixel, The minimum depth and the maximum depth among the depths of the neighboring range pixels are determined. The display position of the lower limit designation icon and the upper limit designation icon on the focal length slider S1 may be automatically shifted so as to correspond to the determined minimum depth and maximum depth.

Further, in the emphasized image generation processing, when the user taps an arbitrary position on the image and focuses a pixel at the same focal distance as the position, and generates an emphasized image in which pixels at different focal distances are not focused , A display may be performed to distinguish a pixel at the same focal distance from a pixel at a different focal distance so that a pixel can be focused (or non-focused) to the user.

Specifically, the processing section 10 determines pixels included in the tab position by the user. In this case, for example, the processing section 10 specifies a pixel located at the center of the tap position, and sets the pixel as a processing target pixel. Then, the processing section 10 refers to the depth map, specifies the pixels having the same focal distance as the pixel to be processed, flash-displays the pixels, and makes the user distinguishable from other pixels.

Fig. 28 is a diagram showing an example of an image distinguishing display in this case, and shows an example of an output image in which a doll of a rabbit is displayed as a foreground in the center and a tree or a building as a background.

As shown in Fig. 28 (1), when the user taps the fuselage portion of the puppet of the rabbit as the foreground, the pixel corresponding to the tab position is judged, and the pixel at the same focal distance is displayed in flash. Assuming that the focal lengths of the pixels constituting the puppet of the rabbit are all the same in order to simplify the explanation, as shown in (1) of Fig. 28, the pixels constituting the puppet of the rabbit become objects of discrimination display, The image portion of the doll of FIG.

28 (2), when the user taps a part of the background tree, the pixel corresponding to the tab position is judged, and the pixels at the same focal distance are displayed in flash. Assuming that the focal lengths of the pixels constituting the background or the pixels constituting the background are all the same for the sake of simplicity of explanation, pixels constituting the background tree are the objects of discrimination display, as shown in Fig. 28 (2) , The image portion of the background is displayed in flash.

However, as the above distinguishing display, pixels having the same focal distance may be displayed in a translucent manner in a color different from that of the other pixels, or may be filled in with different colors, or may be displayed in a predetermined shape or hatching , It is also possible to realize display of distinction of pixels at the same focal distance.

7. Seventh Embodiment

In the above embodiment, a certain time is required until an emphasized image is generated in the emphasized image generation processing after the imaging by the imaging section 50 is performed. In the meantime, since it is inefficient to simply wait for the user, in the emphasized image generation processing, a process of designating which image portion is focused / unfocused to the user may be interposed.

Specifically, a plurality of photographed images photographed at different focal lengths are set as input images, and processing for detecting, for example, an edge is performed as a feature of the input image. Then, for each of the input images, an image obtained by multiplying the input image and the edge image generated for the input image by a pixel unit (hereinafter, referred to as &quot; ), And displays the images as a reference image in a sequential order (or in a far order) in which the focal distances are close to each other. Repeat the display if it has been done one time.

The user browses an enemy image sequentially displayed and performs a selection operation (e.g., tap operation) for selecting the image when an enemy image in which the subject to be focused is clearly displayed is displayed. In this case, an input image corresponding to the selected image is specified by the user's selection operation, an image in which the pixel at the focal distance of the input image is focused, and other pixels are not focused is generated and displayed.

Fig. 29 is a flowchart showing an example of the flow of the fourth enhanced image generation processing executed by the processing section 10 in place of the enhanced image generation processing of Fig. 8 in this case.

First, the processing section 10 performs an enemy image generation process (step K1). Specifically, for each of a plurality of input images, a product image is calculated by calculating a product of the input image and an edge image obtained for the input image in the edge detection process of step A11.

However, since the brightness of an image of an input image and an edge image may be lowered, it is preferable to perform a process of raising the brightness of the generated red image.

Next, the display control section 127 performs control to display the first enemy image on the display section 30 (step K3). Specifically, for example, an enemy image corresponding to the input image having the shortest focal distance or an enemy image corresponding to the input image having the longest focal distance is displayed on the display section 30. [

Then, the processing section 10 judges whether or not a tap operation has been performed on the touch panel 21 (step K5). If it is determined that the tap operation has not been performed (step K5; No) It is determined whether or not a predetermined time (for example, one minute) has elapsed (step K7).

If it is determined that the predetermined time has not elapsed (step K7; No), the display control section 127 controls the display section 30 to display the next enemy image (step K9). Then, the processing section 10 returns the process to step K5.

On the other hand, if it is determined that the tap operation has been performed (step K5; Yes), the processing section 10 specifies the focal length of the input image corresponding to the red image (step K11). Then, the enhanced image generation unit 123 generates an enhanced image (step K13). Specifically, when the depth map is stored in the storage unit 80 in which the depth map has been generated, based on the depth map stored in the storage unit 80, the pixel at the depth corresponding to the focal distance specified in Step K11 And other pixels are subjected to shading processing to generate an emphasized image in a non-focused state. If the depth map is not in a state in which the generation of the depth map is completed, the process waits until the depth map is generated.

Next, the display control section 127 controls the display section 30 to display the generated enhanced image (step K15). Then, the processing unit 10 ends the fourth enhanced image generation processing.

On the other hand, if it is determined in step K7 that the predetermined time has elapsed (step K7; Yes), the processing section 10 specifies the focal distance of the default input image (step K17). Here, the default input image may be, for example, an input image having the shortest focal length or an input image having the longest focal length. Then, the processing section 10 shifts the processing to step K13.

According to the above processing, it is difficult to automatically determine the position to focus on because of the user's preference of which focal distance is to be focussed. However, by specifying the position to focus on the user during execution of the enhanced image generation processing, Can be solved. In addition, it is possible to solve the inefficiency of simply waiting for the user during execution of the enhanced image generation processing.

In step K3, the display control unit 127 sequentially displays a plurality of input images picked up by the image pickup unit 50 as reference images, for example, . In addition, an image such as a feature detection image in which a feature (e.g., an edge) is detected for each of a plurality of input images picked up by the image pickup section 50, or a contrast-corrected image in which contrast correction is performed for each of a plurality of input images, And may be sequentially displayed as a reference image.

8. Eighth Embodiment

The eighth embodiment assumes a case where highlight clipping occurs in the input image and a shading process (hereinafter referred to as &quot; highlight clipping effect shading process &quot;) for effecting the highlight clipping area Fig. In the area including the highlight clipping, it is preferable to perform the shading so as to expand in response to the light intensity. However, when a shading process is performed on an input image including a highlight clipping area using a filter that performs a simple convolution operation (spatially weighted average), an image of an unnatural shading result that is different from the real world is obtained .

Here, the normal shading processing is given by the following equation (5).

(X) is the pixel value of the pixel x of the shading object image, "I (x)" is the pixel value of the pixel x of the shading image, "Denotes a spatial weight (shading kernel), and" n "denotes a pixel near the pixel x.

For the highlight clipping area, the intensity of the original light becomes unclear. The seaweed value represented by 8 bits is saturated and the seawater value of the pixel whose original seawater value is 256 or more is corrected to 255. [ That is, when the intensity of the light in the pixel x is taken as "L (x)", the pixel value is calculated as "I (x) = min (L (x), 255)". Due to this, when the shading process is performed using the equation (5), the weighted average is unduly low. As a result, no expansion occurs in the highlight clipping area.

Here, in the present embodiment, shading processing is performed according to the following expressions (6) and (7).

That is, instead of performing the weighted average using the saturated pixel values, a weighted average of the light intensity is performed, and the pixel value of the input image is corrected using the intensity of the light obtained by the weighted average, Get the value.

As another technique for avoiding the above problem, it is possible to consider weighted averaging of pixel values by setting a large weight for the highlight clipping area. Specifically, for example, a weighted average is performed according to the following equation (8).

In Equation (8), "w2" is a weight for the highlight clipping area, and a value that is larger as the pixel value included in the highlight clipping area can be set. Specifically, for example, the weight "w2" can be calculated and set according to either one of the following equations (9) to (11).

However, "s" in the expression (11) is an appropriate parameter value.

32 is a flowchart showing an example of the flow of the highlight clipping effect shading processing executed by the processing section 10 in this case.

First, the processing section 10 inputs an image to be processed which is an image to be processed (step L1). Then, the processing unit 10 performs a highlight clipping area detecting process for detecting a highlight clipping area from the input image to be processed (step L3).

Next, the processing unit 10 sets the weight "w2" for the highlight clipping area on a pixel-by-pixel basis (step L5). More specifically, the weight "w2" for the highlight clipping area is calculated and set for each pixel according to either one of the expressions (9) to (11).

Next, the processing section 10 uses the weight "w2" for the highlight clipping area set in step L5 and the pixel value "I" of the pixel to be processed, Weighted average calculation is performed for the value "I" to acquire an image of the shading result composed of the modified pixel value "I" (step L7). Then, the processing unit 10 ends the highlight clipping effect shading process.

However, in order to perform more realistic shading in the highlight clipping effect shading process, a model based on the assumption that the intensity of light becomes larger (closer to the center) of the highlight clipping area is applied to perform the shading process good. For example, the weight "w2" for the highlight clipping area may be set large corresponding to the distance from the boundary between the highlight clipping area and the non-highlight clipping area. More specifically, a process of detecting a boundary between a highlight clipping area and a non-highlight clipping area is performed by using a known boundary detection method, and a distance from the detected boundary to a pixel included in the highlight clipping area is long W2 "for the highlight clipping area may be set to a larger value and the weighted average calculation for the pixel value" I " may be performed.

In addition to the above highlight clipping effect shading processing, for example, a real pixel value of the highlight clipping area is estimated using a low-exposure image of the same angle of view as the reference image, and the shading processing .

For estimating the real pixel value of the highlight clipping area, various techniques can be used. Specifically, for example, the pixel values of two corresponding points (except for the highlight clipping area) of two images of the reference image and the low exposure image on the basis of the shading are defined as "y = a x ^quot ; ^{b &} quot ;, and the value of the exposure conversion parameter (a, b) included in the exposure conversion function is calculated by the minimum multiplying method. Then, by using the calculated values of the exposure conversion parameters (a, b), the pixel value of the low-exposure image is converted to estimate the real pixel value of the highlight clipping area.

The various shading processes in consideration of highlight clipping may be performed by, for example, a plurality of shot images (input images) picked up by a focus bracket method for obtaining a plurality of images by shifting the focus positions as described in the first or second embodiment Image) or a plurality of captured images (input images) captured by a parallax method as described in the third embodiment can be applied as an image to be processed.

That is, as the subroutine of the image processing program for executing the image processing described in the above various embodiments, let the storage section 80 store the highlight clipping effect shading processing program for executing the highlight clipping effect shading processing And the processing section 10 reads and executes the highlight clipping effect shading processing program in the image processing executed in accordance with the image processing program, thereby realizing effective shading considering the highlight clipping with respect to the image picked up by each method .

9. Variations

It is needless to say that the embodiments to which the present invention is applicable are not limited to the above-described embodiments, but can be appropriately changed within a range not departing from the gist of the present invention. Hereinafter, modified examples will be described, and the same reference numerals are assigned to the same components as those of the above-described embodiment, and repetitive description thereof will be omitted.

9-1. Image processing apparatus

In the above-described embodiment, the embodiments in which the image processing apparatus is a smart phone have been described. However, the electronic apparatuses applicable to the image processing apparatus are not limited to smart phones. For example, an electronic apparatus such as a camera, a digital camera, a tablet terminal, or a PDA may be applied as an image processing apparatus. In the case of an electronic apparatus having an image pickup section, it is possible to apply the image processing apparatus of the present invention to various electronic apparatuses.

Further, for example, in the image processing apparatus, the user selects the depth based on the depth map or the foreground image displayed on the display unit, and displays the enhanced image generated based on the selected depth on the display unit of the predetermined information processing apparatus Maybe. More specifically, for example, the user communicates with a personal computer held in his or her hand with a personal computer. Then, the user selects the depth based on the depth map or the foreground image displayed on the display of the smartphone. The smart phone generates an emphasized image based on the depth selected by the user, and transmits the emphasized image to the personal computer. The personal computer displays the emphasized image received from the smart phone on the display, and the user confirms the emphasized image on the display screen of the personal computer.

In this case, however, the image data may be transmitted from the smart phone to the personal computer, and the personal computer may generate the enhanced image instead of the smart phone and display the enhanced image on the display of the personal computer. In this case, a smart phone and a personal computer become the image processing apparatus of the present invention.

9-2. Feature quantity

In the above-described embodiment, the description has been given taking as an example the case where an edge is detected as a feature from a plurality of input images having different depths of focus by changing the focal length, and the depth is estimated using the edge intensity as a feature amount. However, It is not.

For example, a DoG (Difference of Gaussian) value indicated by the difference of the Gaussian function may be detected as a feature amount having a correlation of focusing and positive (positive) in the same manner as the edge intensity. In this case, for each of the plurality of input images, for example, filter processing using two Gaussian filters having different standard deviations? (Variance? ² ) is performed to obtain two smoothed images having different standard deviations?. Then, a DoG image is obtained by calculating the difference between the two smoothed images for each of the plurality of input images. The pixel value (DoG value) of each pixel of the DoG image thus obtained is used as a feature amount, and an appropriate arithmetic operation process is performed as in the above embodiment to estimate the depth of each pixel. In this case as well, a depth map reflecting the focusing degree can be generated for each pixel.

Further, as the evaluation indexes other than the focusing degree, for example, the background degree of the image is detected as the feature amount, and the detected background degree is used as the feature amount to perform the appropriate arithmetic operation processing as in the above embodiment, A map reflecting the background degree may be obtained. In this case, an input image in which a background is displayed among a plurality of input images is preferentially selected, and a map reflecting the background degree is generated for each pixel.

Other than these, color characteristics such as the ratio of the representative color and the color used in the image, the color distribution, and the like may be detected as the characteristic quantities. Further, the texture, which is a regular pattern included in the image, may be detected as the feature amount. Further, the shape of the object in the image may be detected as the feature amount. Further, specific azimuth information may be detected as a feature amount by using the orientation selectivity filter. Either way, if the feature amount is an index of the focusing degree, a map reflecting the focusing degree is obtained, and if the feature amount is another evaluation index, a map reflecting the evaluation index is obtained.

9-3. Appropriate arithmetic processing

In the above-described embodiment, as the appropriate arithmetic processing, the score value is calculated using the energy function of Equation (1), and the input image having the maximum score value is determined as the focused input image. (12), instead of (1), the score value may be calculated and the input image with the smallest score value may be determined as the focused input image.

In Expression (12), each term on the right side of Expression (1) is not a minus but a plus. In this formula, a cost value corresponding to the edge intensity of the corresponding edge image of the target pixel is added to the target pixel, and a first input image that is focused on the target pixel, The penalty value is added to the score value when the second input image that is focused on the pixel near the pixel is different. Then, an input image whose score value satisfies a predetermined low value condition is determined as the focused image.

The equation corresponding to the equation (2) in which the energy function of the equation (12) is extended in one dimension is expressed by the following equation (13).

When the energy function of the equation (1) is used, the input image having the maximum score value is determined as the focused input image. However, this technique is merely an example. For example, when the number of input images is N, for example, a score value of a predetermined number (for example, 3%) is specified in ascending order of the score value, and one of the score values is selected, It may be determined. That is, it is not necessary to select the maximum score value, and the second or third largest score value may be selected. This is an example showing that an input image whose score value satisfies a predetermined high value condition is determined as an input image that is focused.

Similarly, when the energy function of the equation (12) is used, a score value of a predetermined number (for example, 3) may be specified in the order of smaller score values, and one score value may be selected from among them. This is an example showing that an input image whose score value satisfies a predetermined low value condition is determined as an input image that is focused.

9-4. Display image

In the above-described embodiment, the foreground image generated based on the depth map is displayed on the display unit 30 as the display image. However, the foreground image must be displayed on the display unit 30 .

More specifically, instead of the fore-focus image, a representative image selected from a plurality of input images may be displayed on the display unit 30 as a display image. In this case, for example, the input image picked up at the earliest time, the input image picked up last, and the input image whose focal distance is the closest to the middle value can be selected as representative images.

Alternatively, the depth map may be displayed on the display unit 30 as a display image, or a synthesized image obtained by synthesizing an arbitrary input image among a plurality of input images may be displayed on the display unit 30 as a display image .

Either way, the user operates the operation unit 20 while viewing the above-mentioned display image displayed on the display unit 30, and selects a portion to be corrected in the image.

9-5. Display of depth selection window

The display control unit 127 controls the display unit 30 to display a predetermined display image and controls to display a depth selection operation window for allowing the user to select the depth at a predetermined position of the display screen . In this case, the user selects the depth from the depth selection operation window through the input unit 20.

Fig. 17 is a diagram showing an example of a display screen in this case, and shows a state in which a foreground image is displayed on the display unit 30. Fig. However, only the puppet part of the rabbit is shown in the foreground, and the part of the background tree is omitted. In this figure, a depth selection operation window W1 for the user to select the depth is displayed at the lower left corner of the display screen. The depth selection operation window W1 is represented by a vertically long gauge divided by the depth and the user is able to select the depth by moving the depth selection bar BAR displayed at the center of the gauge by the tab operation .

18 is a flowchart showing the flow of a second enhanced image generation process executed by the enhanced image generation unit 123 in this case. It should be noted that the same steps as those in the emphasized image generating process shown in Fig. 8 are denoted by the same reference numerals and repeated explanation is omitted.

The emphasis-image generating unit 123 first determines a user operation (step H1). If it is determined that the user operation is a tap operation of an image on the display screen (step H1; tap operation of the image), the enhanced image generation unit 123 determines the tapped area (tap area) as the designated area (step H2). Then, the emphasized image generating unit 123 shifts the processing to step D3.

On the other hand, if the user operation is determined to be an operation on the icon of the depth selection operation window (step H1; icon operation of the depth selection operation window), the enhanced image generation unit 123 sets the depth of the moving destination Depth selected by the icon) (step H3). Then, the emphasized-image generating unit 123 refers to the depth map to specify a pixel corresponding to the determined depth (step H5). Then, the enhanced image generation unit 123 generates an enhanced image based on the specific pixel (step H7). More specifically, a process of shading the entire foreground image with the degree of the first shading is performed using a shading filter. Then, in step H5, the pixel value of a specific pixel is synthesized with the shading image obtained as a result of the shading processing to generate an enhanced image.

9-6. Handling shading again

In the emphasized image generation processing described in the above embodiment, it is determined whether or not the area on the image designated by the user is a shaded area in the case where the user operation is detected. If the determination result is affirmative, The degree of shading of the area may be set to a degree of shading different from that of the predetermined degree of shading, and the enhanced image may be shaded again.

Specifically, when the user taps a predetermined portion of the shaded portion in the emphasized image generation processing through the touch panel 21, the processing portion 10 determines whether the portion is already shaded , &Quot; shading completion portion &quot;). If it is determined that the shading is completed, the processing unit 10 makes a confirmation notification to confirm the user whether the degree of shading of the shaded portion is strong or weak. When it is selected by the user to weaken the degree of shading, the processing unit 10 performs the shading processing with respect to the portion where the shading is completed by making the degree of shading smaller than the specified degree of shading. On the other hand, when it is selected by the user to intensify the degree of shading, the processing section 10 performs the shading processing with respect to the portion where the shading is completed by making the degree of shading larger than the degree of the specified shading. Then, control is performed so that the emphasized image resulting from the shading processing is displayed on the display unit 20. [

9-7. User manipulation

In the above-described embodiment, the operation when the user creates the enhanced image or the modification of the enhanced image is described as the tap operation or the drag operation with respect to the touch panel 21, but the user operation to the touch panel 21 It is not necessary that the generation of the emphasized image and the correction of the emphasized image are performed in accordance with the above.

For example, when the image processing apparatus includes an operation button (including a cross key), an emphasized image may be generated or an emphasized image may be corrected in accordance with the operation of the operation button. In this case, for example, a designation icon (e.g., an icon of an arrow) for designating the position on the image in a state in which the front-focus image is displayed on the display unit 30 is displayed, The icon may be moved to designate the portion to be emphasized in the display image. In addition, in the state in which the emphasized image is displayed on the display unit 30, the designation icon for designating the position on the image is displayed, and the user moves the designation icon by using the operation button, You may specify the part you want.

9-8. Creation of a highlight image

In the above-described embodiment, the emphasis image is generated based on the fore-focus image. Specifically, an enhanced image is generated by combining pixel values of an input image having a depth of a portion to be emphasized with an image obtained by performing shading processing on a foreground image. However, it is also possible to generate an enhanced image using only the input image, without generating a foreground image.

More specifically, for example, when an operation input is made by the user, the pixel value of the target pixel in the input image at the depth to be emphasized by the operation input is added to the input image other than that depth The shading processing may be performed on pixels other than the target pixel in the original image to generate an enhanced image. It is also possible to combine the pixel values of the input image of other depths with the pixel value of the input image of the corresponding depth of the selected pixel in the designated area designated by the user operation to generate the enhanced image. That is, the pixel value of the target pixel in the input image at a depth to be emphasized by the operation input of the user may be combined with the input image corresponding to the depth other than the depth to generate the enhanced image.

9-9. Emphasis image generation timing

Instead of generating the emphasized image when the user performs the tab operation or the drag operation, the emphasized image may be generated by triggering any other action. Specifically, for example, a button for instructing generation of an emphasized image is displayed on the display section 30, and an emphasized image may be generated when the push button is pressed through the touch panel 21 . Further, in the case of an image processing apparatus having an operation button, it is also possible to detect the depression of the operation button and generate an enhanced image.

Further, when a sensor (for example, an inertial sensor such as an acceleration sensor or a gyro sensor) for detecting vibration is incorporated in the image processing apparatus and an operation of shaking the image processing apparatus by the user is detected, .

Further, a microphone may be provided in the image processing apparatus as a sound collecting section, and an emphasized image may be generated when a predetermined voice is input to the microphone. In this case, the image processing device stores the audio data obtained by digitizing several kinds of voices triggered by the generation of the enhanced image, and executes a speech recognition process of collating the audio inputted from the microcomputer with the stored audio data, It may be determined whether or not the generation of the image is instructed by the user. In these cases, for example, depths to be emphasized are sequentially changed from shallow depths of an input image to a shallow depth or from a deep depth of an input image to a shallow depth, sequentially generating an emphasized image, Images can be sequentially displayed, and the like.

9-10. Pinch manipulation

When the user performs a pinch operation on the touch panel 21, the display image may be enlarged or reduced or rotated in accordance with the operation of the pinch. Specifically, when it is detected that the user has touched the touch panel 21 with two fingers and an operation of increasing / decreasing the distance between the two points of the taps is detected, for example, A pixel on the depth map corresponding to the center coordinate may be specified and the entire display image may be enlarged / reduced based on the specified pixel. Further, when it is detected by the user that an operation of turning a finger to change the angle of a straight line connecting two points is performed, a pixel on the depth map corresponding to the center coordinate between the tapped two points is specified , The whole of the display image may be rotated on the basis of the specific pixel.

9-11. Re-estimation of depth

9, the penalty value is changed based on the degree of cosine similarity between the drag direction and the edge direction, the score value calculation process is performed on the pixels in the designated area, and the depth Re-estimated. That is, the second-type parameter value is changed for the pixels in the designated area, and the score value calculation process is executed. On the other hand, however, the cost value calculation processing may be performed by changing the cost value for the pixels within the designated area. That is, the first-type parameter value may be changed for the pixels in the designated area to execute the score value calculation process.

In the enhanced image modification processing, it is necessary to generate a modified image in which the edge is emphasized in the designated area designated by the user. In order to emphasize an edge, it is necessary to select a pixel corresponding to an edge from an input image having a depth close to the depth of a representative pixel in the designated area. Since the cost value is calculated as the inverse number of the edge strength, the cost value becomes smaller as the edge strength becomes stronger. That is, an image with a strong edge strength has a smaller cost value. Therefore, in order to select a pixel corresponding to an edge from an input image having a near depth, it is preferable to reduce a cost value for an input image having a depth close to the depth of a representative pixel in the designated area. The cost value can be independently changed for each input image. As a result, it is possible to make the input image easier to select by reducing the cost value of only one input image, or conversely, it is possible to make it difficult to select the input image by increasing the cost value of only one input image.

It is also possible to re-estimate the depth by changing only one of the cost value and the penalty value. However, it is also possible to re-estimate the depth by changing both the cost value and the penalty value. That is, it is possible to re-estimate the depth by changing the at least one of the first-type parameter value and the second-type parameter value and performing the score value calculation process.

9-12. Penalty value

The value that can be set as the penalty value is not limited to two, the first penalty value and the second penalty value. For example, the third penalty value 897 may be set in addition to the first penalty value 893 and the second penalty value 895 as shown in the penalty value table 89 in Fig. 16 . More specifically, the third penalty value 897 may be set as the penalty value when the input image that is focused on the target pixel and the neighboring pixels is changed to "3 or more" in the image number 833b . The penalty value table 895 of Fig. 16, the "normal time" of the type 891, the can is "P _3" defined as a penalty value (897) of the three. &Quot; P ₁ &lt; P ₂ &lt; P _{3 &} quot ;.

In addition, in the case of &quot; correcting the emphasized image &quot; of the category 891, there is a case where the drag direction and the edge direction by the user satisfy a predetermined approximation condition and the case where the drag direction by the user and the edge direction satisfy a predetermined divergence condition In the case, another penalty value is fixed. Specifically, for the former, "P ₃ -β" is defined as the third penalty value 897. In the latter case, "P ₃ + β" is determined as the third penalty value 897. However, the value of &quot;?&Quot; can be appropriately set, for example, &quot;? = 1 &quot;.

Further, the penalty value may be changed corresponding to the relative relationship between the drag direction of the user and the edge direction in the image. More specifically, for example, a penalty value is individually determined corresponding to the degree of similarity between the drag direction and the edge direction (hereinafter, referred to as &quot; penalty value by direction similarity degree &quot;), Processing may be performed. As an index value indicating the degree of similarity between the drag direction and the edge direction, similar to the above-described embodiment, a cosine similarity degree or a cosine distance can be applied.

Specifically, instead of the penalty value P (p) defined in the above embodiment, a penalty value P _dir (p) by direction similarity degree is set. For example, a value that is the same as the penalty value at the normal time is set as the penalty value by direction similarity degree when the cosine similarity degree is "1" (parallel between the drag direction and the edge direction). In addition, the penalty value is determined so that the value gradually increases from the penalty value at the normal time as the cosine similarity becomes smaller (as the drag direction and the edge direction become distant from each other), and the cosine similarity is set to "0" (the drag direction and the edge direction are orthogonal ), A predetermined maximum value is set as the penalty value. Then, the cosine similarity degree between the drag direction and the edge direction of the user is calculated, and the arithmetic processing is appropriately performed using the penalty value by the determined direction similarity degree with respect to the calculated cosine similarity degree.

The penalty value for each direction similarity degree is useful when re-estimating the depth described in the above embodiments. As described above, in the present invention, the closer the drag direction and the edge direction are, the smaller the penalty value. On the contrary, the penalty value increases as the drag direction and the edge direction become different from each other. When the user finds that a display defect, that is, a calculation error of depth, is occurring in the display image, a point where a calculation error occurs is specified by a drag operation, and a correction request for the emphasized image is made.

When the user drags the object so that it follows the edge direction of the object correctly displayed, the cosine similarity degree between the drag direction and the edge direction becomes a value close to &quot; 1 &quot;. In this case, the arithmetic processing is appropriately performed using the penalty value for each direction similarity degree in which the cosine similarity is set to &quot; 1 &quot;. Since the penalty value in this case is the same value as the normal penalty value, The depth map generated at the normal time can be saved.

On the other hand, in the case where the user has dragged to follow a portion where a calculation error has occurred, as long as the edge direction of the portion where the calculation error occurs is not the same as the edge direction of the portion correctly displayed, A difference is generated, and the cosine similarity degree becomes smaller than "1". For example, when the drag direction and the edge direction are orthogonal, the cosine similarity becomes &quot; 0 &quot;. In this case, the arithmetic processing is appropriately performed using the penalty value by direction similarity degree when the cosine similarity degree is &quot; 0 &quot;. Since the maximum value is set as the penalty value in this case, the penalty becomes strongest. As a result, the depth map showing the depth more accurately with respect to the portion where the calculation error occurs is regenerated. By modifying the emphasized image using the depth map thus reproduced, it is possible to appropriately correct the portion where the calculation error occurs.

Further, the penalty value from one input image to another input image may be determined individually. Specifically, instead of the penalty value P (p) defined in the above embodiment, the inter-image penalty value P _ij (p) may be set. Note that &quot; i &quot; and &quot; j &quot;

For example, it is assumed that the focal length of the input image is "f1 to f4", and the magnitude relation is "f1 <f2 <<<<< f3 <f4". In this case, since "f1" and "f2" are not so different from each other, the inter-image penalty value P ₁₂ (p) between the input image of "f1" and the input image of "f2"Quot; f2 &quot; and the input image of &quot; f2 &quot; are focussed. The inter-image penalty value P ₃₄ (p) between the input image of "f3" and the input image of "f4" is made larger so that "f3" And the input image of &quot; f4 &quot; are focussed.

Since the inter-image penalty value P ₂₃ (p) between the input image of "f2" and the input image of "f3" is made small, the input of "f2"Quot; f3 &quot; becomes difficult to be mutually selected as an input image in which the image and the input image of &quot; f3 &quot; are focused. This makes it possible to increase the penalty for the input image of "f2" from the input image of "f1" or the input image of "f4" of the input image of "f3"Quot; f3 &quot; from the input image of &quot; f2 &quot;, it is possible to lower the penalty and make it difficult for the input image to be selected as an image far away from the focal length.

9-13. Anti-shake measures

(Hereinafter, referred to as &quot; subject shake area &quot;) in the input image and to make it difficult for the estimated depth (the input image selected as being focused) to change in the subject shake area, Or at least one of the values may be changed. A subject shake is a phenomenon in which a subject moves during shooting and a moving speed thereof is faster than a shutter speed of the camera, thereby causing a subject in the image to shake. Detection of a subject shake area can be realized by a conventionally known technique.

If the subject shake area is detected, the score value calculation process is performed by changing the penalty value P (p) in the subject shake area to a value larger than the normal penalty value, or the score value C (p, _Rp ) To a value smaller than the cost value calculated by using the edge strength, and performs the score value calculation processing. This makes it possible to generate a depth map with an appropriate depth for the subject shake area.

9-14. Recording medium

In the above embodiment, various programs and data for image processing are stored in the storage unit 80 of the image processing apparatus, and the processing unit 10 reads and executes these programs, The image processing in the first embodiment is realized. In this case, the storage unit 80 may be a memory card, a compact flash (CF) card, a memory stick, a USB memory, a CD-RW (optical disk), a hard disk, (Recording medium, external storage device) such as an MO (magneto-optical disk), or the like, and the various programs and data described above may be stored in these recording media.

30 is a diagram showing an example of a recording medium in this case.

The image processing apparatus is provided with a card slot 710 for inserting the memory card 7 and reads the information stored in the memory card 7 inserted in the card slot 710 or reads the information stored in the memory card 7 And a card reader / writer (R / W) 720 for writing information. The card reader writer 720 performs an operation of writing the program and data recorded in the storage unit 80 to the memory card 7 under the control of the processing unit 10. [ By reading the program or data recorded in the memory card 7 from an information processing apparatus (for example, a personal computer) other than the image processing apparatus, it is possible to realize the image processing in the above- It is possible.

1: Smartphone
7: Memory card
10:
20:
21: Touch panel
30:
40:
50:
55: Antenna for mobile phone
60: Wireless communication circuit for mobile phone
70: Watch section
75: inertia sensor
80:

Claims (18)

Feature detecting means for detecting a predetermined feature from a plurality of input images;
And a depth map generating means for generating a depth map based on the detection result of the feature detecting means
And the image processing apparatus. The method according to claim 1,
Characterized in that the feature detecting means includes a feature amount detecting means for detecting a predetermined feature amount from the plurality of input images,
And depth estimation means for estimating a depth, which is information indicating which of the plurality of input images is focused on the target pixel, using the feature amount detected by the feature amount detection means ,
Wherein the depth map generating means generates the depth map based on the estimation result of the depth estimation means. The method of claim 2,
Wherein the depth estimation means calculates a depth parameter value for each of the plurality of input images detected by the feature amount detection means by using a feature amount of each of the plurality of input images, Using a second type parameter value determined as a penalty in the case where a first input image that is focused on the target pixel and a second input image that is focused on a neighboring pixel of the target pixel are different from each other, And determines an input image that has been focused on the target pixel. The method according to claim 2 or 3,
Wherein the feature amount detection means detects the feature amount from a plurality of input images or a reduced input image obtained by reducing the plurality of input images,
Wherein the depth map generation means generates a reduced depth map based on the estimation result of the depth estimation means,
Restoration means for restoring the reduced depth map using a predetermined restoration method,
And correction means for generating a depth map corrected by using the predetermined correction image with the depth map restored by the restoration means. The method according to claim 1,
Characterized in that the feature detection means has a feature point detection means for detecting a feature point from the input image by using the captured images captured at a plurality of viewpoints (viewpoints) at which parallax occurs,
A viewpoint positional relationship estimation unit that estimates a relative positional relationship between viewpoints of the plurality of input images based on the feature points detected by the feature point detection unit;
A pixel position estimation means for estimating a three-dimensional position of a pixel constituting the plurality of input images by performing a predetermined estimation operation,
Further comprising:
Wherein the depth map generation means generates the depth map based on the estimation result of the viewpoint position relation estimation means and the estimation result of the pixel position estimation means. The method of claim 5,
Display means,
Display control means for performing control to display, on the display means, a guide display for guiding the user to a method of moving a magnetic device for generating parallax in the plurality of shot images;
A movement amount detecting means for detecting a movement amount (amount)
Further comprising:
Wherein the display control means changes the display of the guide display based on the detection result of the movement amount detecting means. The method according to any one of claims 1 to 6,
Display means,
An input means for accepting a user operation,
A composite image obtained by synthesizing a pixel value of a pixel in an input image corresponding to a specific depth to an input image corresponding to a depth other than the specific depth is generated based on the depth map in correspondence with a user operation to the input means Synthesized image generating means
Further comprising:
Wherein the display means displays the composite image generated by the composite image generation means. The method according to claim 1,
Display means,
An input means for accepting a user operation,
Focus image generating means for generating a full focus image using the depth map and the plurality of input images;
And a shading processing means for performing a shading process on the foreground image so that a pixel value of a pixel in an input image corresponding to a specific depth based on the depth map corresponding to a user operation on the input means is subjected to the shading processing And a synthesized image generation means
Further comprising:
Wherein the display means displays the composite image generated by the composite image generation means. The method of claim 8,
Further comprising display control means for causing the display means to display a designation figure for designating the focusing target area, which is an area to be focused, corresponding to the user operation on the input means,
Wherein the synthesized image generating means generates the synthesized image with the depth of the pixel included in the designation figure among the input images set to the specific depth based on the depth map. The method according to claim 7 or 8,
Wherein the input means has a touch panel integrally formed with the display means,
Wherein the display means displays a confirmation screen enabling confirmation of an image pickup object,
Storage means for storing a focusing target position, which is a position on a confirmation screen to be focused, corresponding to a tap operation on the touch panel on the confirmation screen;
An image pickup means for picking up an image corresponding to a predetermined image pickup condition after the focusing target position is stored in the storage means,
Further comprising:
The feature detecting means detects the feature by using a plurality of captured images captured by the image capturing means as the plurality of input images,
Wherein the composite image generating means generates the composite image with the depth corresponding to the focusing target position stored in the storage means at the specific depth. The method according to claim 7 or 8,
Wherein the input means has a touch panel integrally formed with the display means,
Wherein the display means displays a predetermined display image,
Focusing area setting means for setting a focusing subject area that is an area to be focused, corresponding to a pinch operation on the touch panel on the display screen on which the display image is displayed;
Shading processing means for performing shading processing on a region other than a focusing target region set by the focusing target region setting means to a predetermined degree of shading
Further comprising:
Wherein the shading processing means includes shading degree changing means for changing the shading degree in accordance with a pinch operation on the focusing target area set by the focusing target area setting means. The method according to any one of claims 1 to 6,
Wherein the plurality of input images are a plurality of captured images captured at different focal lengths,
Display means,
An input means for accepting a user operation,
Display control means for performing control to display a focal length range designation image for designating a focal length range by a user at a predetermined position of the display means;
A synthesizing unit configured to focus a pixel of a depth corresponding to a focal length range specified by a user operation on the image for specifying the focal length range and to generate a composite image in which other pixels are non- Generating means
Further comprising:
Wherein the display means displays the composite image generated by the composite image generation means. The method according to any one of claims 1 to 6,
Wherein the plurality of input images are a plurality of captured images captured at different focal lengths,
Display means,
An input means for accepting a user operation,
Display control means for performing control to sequentially display a plurality of reference images based on the plurality of input images on the display means,
And a synthesizing unit configured to perform, on the basis of the depth map, a pixel having a depth corresponding to the focal length of the input image corresponding to the reference image designated by the user, The synthesized image generating means
Further comprising:
Wherein the display means displays the composite image generated by the composite image generation means. The method according to claim 7 or 8,
Wherein the input means has a touch panel integrally formed with the display means,
Wherein the synthesized image generating means dynamically changes the specific depth based on a drag operation on the touch panel to dynamically generate the synthesized image,
Wherein the display means displays the composite image dynamically generated by the composite image generation means in accordance with the drag operation. The method according to any one of claims 2 to 4,
Further comprising input means for accepting a user operation,
The depth estimating means has depth re-estimating means for re-estimating the depth with respect to pixels included in the designated region specified by the user operation,
Wherein the depth map generating means includes depth map regenerating means for regenerating a depth map based on a result of re-estimation of the depth by the depth re-estimating means. The method of claim 3,
Further comprising input means for accepting a user operation,
The depth estimating means has depth re-estimating means for re-estimating the depth with respect to pixels included in the designated region specified by the user operation,
The depth map generating means has depth map regenerating means for regenerating the depth map based on the result of re-estimation of the depth by the depth estimator,
Wherein the depth estimating means estimates at least one of the first type parameter value and the second type parameter value with respect to the pixels included in the designation region from a parameter value used in the original depth estimation by the depth estimation means And changing the value to a different value to appropriately perform the arithmetic processing to re-estimate the depth. Detecting a predetermined characteristic from a plurality of input images,
And generating a depth map based on the detection result of the feature. On the computer,
A feature detecting step of detecting a predetermined feature from a plurality of input images;
And a depth map generation step of generating a depth map based on the detection result of the feature detection step.

Images